JP2007188211A - Image processor, image pickup device, image processing method and program therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor capable of reducing noise by a coring processing and also suppressing occurrence of edge blur or phase deviation due to the coring processing. <P>SOLUTION: The micro-amplitude signal of a sub-band image signal including high band signal components transformed by a wavelet transformation processing part 1 is transformed into 0 value by a coring processing part 2, and a sub-band image signal including low band signal components and a sub-band image signal obtained by the coring processing are composed in postive phase or negative phase different combination to restore the two image signals by a wavelet inverse transformation processing part 3, and the two image signals are composed by a predetermined phase, and an edge waveform in an output image signal is output as a rotationally symmetrical edge waveform. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image processing device, an imaging device, and an image processing method for reducing noise of an image signal, and more particularly, a multi-resolution conversion method such as a wavelet transform that converts an image signal into a frequency band image signal separated for each frequency band. The present invention relates to an image processing apparatus, an imaging apparatus, and an image processing method that reduce noise by using.

  Conventionally, as a technique for reducing noise of an image signal, a noise reduction technique using wavelet transform, which is one of multi-resolution conversions, is known (for example, see Patent Document 1).

  In such a noise reduction method, first, based on wavelet transform, the image signal is separated into subbands of a low frequency band including a low frequency signal component and a high frequency band including a high frequency signal component, and the obtained low frequency band By repeating the same processing on the signal including the signal component, a frequency band image signal (hereinafter referred to as a subband image signal) separated into subbands for each frequency band is generated.

  Next, a process for removing a minute amplitude component called a coring process is performed on the subband image signal having a high-frequency signal component among the image signals for each frequency band. Further, the wavelet inverse transform is performed by synthesizing the separated subband image signals including the subband image signals subjected to coring processing.

  By executing such processing, an image signal in which only a small amplitude noise component included in the high frequency signal component is suppressed is reconstructed.

  In particular, according to such a method, the amount of noise reduction can be reduced by selecting a high frequency band for performing coring processing to reduce noise having a specific frequency component or changing the coring processing conditions. This makes it possible to perform fine noise reduction processing.

  In addition, the wavelet transform is composed of only Daubechies bases with high band separation accuracy but complicated filter coefficients, and only two filter coefficients “1” and “−1” with low band separation accuracy. There is a specific technique called the Harr base, but for consumer devices and the like, the Harr base that requires a smaller hardware scale is used.

  FIG. 13 is a block diagram showing a basic configuration of a conventional image processing apparatus that reduces noise based on such a Harr basis. FIG. 14 is a diagram for explaining the operation of the conventional image processing apparatus.

  Here, an example is shown in which one-dimensional wavelet transform is performed on an image signal and the number of times of separation into subbands is two.

  Hereinafter, the number of times of subband separation will be referred to as a level. That is, the input image signal is set to level 0, the result of performing one conversion on this input image signal is called level 1, and the result of further conversion on this result is called level 2.

  In FIG. 13, an input image signal S is supplied to a wavelet transform processing unit 91, and a level 2 wavelet transform process based on the Harr basis is executed.

  Since the wavelet transform processing unit 91 performs level 2 wavelet transform, as shown in FIG. 14A, the input image signal is a sample of time series signals S0, S1, S2, and S3 for each pixel. And are grouped as signals S0 and S1, and signals S2 and S3. As a result of performing the level 2 wavelet transform on such four samples, as shown in FIG. 14B, the high frequency band Wh1 obtained by dividing the level 1 subband from each sample in the first conversion is obtained. A high frequency signal Sh1 that is a band signal component and a low frequency signal that is a low band signal component of the band Wl1 are obtained. Further, in the second conversion, a high-frequency signal Sh2 in the band Wh2 and a low-frequency signal S12 in the band Wl2 obtained by subband division of the level 2 are obtained from the low-frequency signal in the band Wl1.

  In the wavelet transform processing unit 91, first, an input image signal is supplied to a 4-tap shift register configured with a delay element 911. From these taps, signals S0, S1, S2, and S3 arranged in units of pixels as four samples while being shifted in the sampling period are output and supplied to each filter for separation into subbands.

  The wavelet transform processing unit 91 includes a low-pass filter provided with an adder 912 and a high-pass filter provided with a subtracter 913 as such filters. In the low-pass filter, the adder 912 performs filter processing with filter coefficients “1” and “1” based on the Harr base. In the high-pass filter, the subtracter 913 performs filter processing with filter coefficients “1” and “−1” based on the Harr base.

  In order to perform the level 1 wavelet transform, the signals S0 and S1 are supplied to the high-pass filter 1H0 and the low-pass filter 1L0. The low-pass filter 1L1 is a filter for level 2 conversion processing, and is supplied with signals S2 and S3.

  The high-pass filter 1H0 extracts the high-frequency signal Sh1 in the band Wh1 shown in FIG. 14B, and the low-pass filter 1L0 extracts the low-frequency signal in the band Wl1 shown in FIG. 14B.

  In this way, the level 1 wavelet transform is executed.

  Next, in order to perform level 2 wavelet transform, the signal from the low-pass filter 1L0 and the signal from the low-pass filter 1L1 are supplied to the high-pass filter 2H0 and the low-pass filter 2L0.

  The high-pass filter 2H0 extracts the high-frequency signal Sh2 in the band Wh2 shown in FIG. 14B, and the low-pass filter 2L0 extracts the low-frequency signal Sl2 in the band Wl2 shown in FIG. 14B.

  In this way, the level 2 wavelet transform is executed.

  Among the signals extracted for each subband, signals having high-frequency signal components, that is, the high-frequency signal Sh1 and the high-frequency signal Sh2 are supplied to the coring processing unit 92.

  The coring processing unit 92 includes a coring circuit 921 and a coring circuit 922 that perform nonlinear processing on the amplitude of the input signal.

  In FIG. 13, as the coring circuit 921 and the coring circuit 922, coring circuits that perform coring processing called soft coring are shown as an example.

  That is, the coring circuit 921 and the coring circuit 922 output an input signal whose absolute value is equal to or less than the threshold value as a uniform 0 value, and when the absolute value of the input signal exceeds the threshold value, the threshold value is calculated from the absolute value of the input signal. Output the signal of the reduced value.

  FIG. 13 shows an example in which the threshold value Th1 is set in the coring circuit 921 and the threshold value Th2 is set in the coring circuit 922.

  By executing such processing in the coring processing unit 92, a minute amplitude included in the high-frequency signal Sh1 or the high-frequency signal Sh2 is regarded as noise and converted to a zero value, and noise is reduced. .

  The coring-processed high-frequency signal Sh1 and high-frequency signal Sh2 are supplied to the wavelet inverse transform processing unit 93 as a high-frequency signal Sh1 'and a high-frequency signal Sh2'. In addition, the low-frequency signal S12 extracted by the wavelet transform processing unit 91 is also supplied to the wavelet inverse transform processing unit 93.

  Since the wavelet inverse transform processing unit 93 reconstructs the supplied signal for each subband, in FIG. 13, the synthesis filters 1A0 and 2A0 that perform the average value calculation by the adder 931 and the half divider 932 in FIG. have.

  The synthesis filter 2A0 adds the low-frequency signal S12 and the high-frequency signal Sh2 'to reconstruct a signal corresponding to the band Wl1 shown in FIG.

  That is, as shown in FIG. 14C, the low frequency signal Sl1 'corresponding to the high frequency signal Sh1 of level 1 is restored.

  Next, the synthesis filter 1A0 adds the restored low-frequency signal Sl1 ′ and high-frequency signal Sh1 ′, thereby reducing the high-frequency noise component from the input signal S0 as shown in FIG. 14C. The signal S0 ′ thus restored is restored.

When the above configuration is considered as a transfer function based on the Z transformation, for example, the transfer function from the input to the output of the high-pass filter 2H0 is (1 + Z ⁻¹ −Z ⁻² −Z ⁻³ ), The transfer function up to the output of the low-pass filter 2L0 is (1 + Z ⁻¹ + Z ⁻² + Z ⁻³ ).

Therefore, when each is synthesized by the synthesis filter 2A0, when the coring process is ignored, the transfer function up to the output of the synthesis filter 2A0 is (1 + Z ⁻¹ ), which is equal to the transfer function up to the output of the low-pass filter 1L0. Become.

Further, when the synthesis filter 1A0 synthesizes the transfer function (1 + Z ⁻¹ ) up to the output of the synthesis filter 2A0 and the transfer function (1−Z ⁻¹ ) up to the output of the high-pass filter 1H0, the synthesis filter 1A0 The transfer function up to the output is (1).

  That is, when the coring process is ignored, a signal equal to the input signal S0 is obtained at the output.

Further, for example, synthesis filter 1A0 is possible to perform subtraction processing in, the composite transfer function from the transfer function to the output of the filter 2A0 (1 + Z ^-1) to the output of the high pass filter 1H0 (1-Z ^-1) Is subtracted and the transfer function up to the output of the synthesis filter 1A0 becomes (Z ⁻¹ ), and it is possible to output a signal equal to the signal S1.

  In this way, when the wavelet inverse transform processing unit 93 synthesizes the low frequency signal Sl2 and each high frequency signal, the phase of each high frequency signal is set to normal phase or reverse phase, and the desired input signal is changed by changing the combination thereof. Can be restored.

As described above, the noise of the high frequency signal component superimposed on the image signal can be reduced by the conventional image processing apparatus shown in FIG.
JP 2003-134352 A

  However, in the conventional image processing apparatus, since coring processing is performed even in the vicinity of an image edge including a high frequency component, the symmetry of the edge is lost, and image processing is performed on, for example, an RGB (Red-Green-Blue) signal. In such a case, there is a problem that color misregistration or the like occurs.

  FIG. 15 is a diagram for explaining edge asymmetry that occurs in such a conventional image processing apparatus. Hereinafter, problems in the conventional image processing apparatus will be described with reference to FIG.

  FIG. 15A is a diagram showing waveforms of the original signal S0 near the edge in the image signal and the output signal from each filter of the wavelet transform processing unit 91 shown in FIG.

  As shown in FIG. 15 (a), the low-pass filter 1L0 outputs a low-frequency signal S11 having a sharp edge with respect to the original signal S0, and the low-pass filter 2L0 further outputs an original signal S0. A low frequency signal Sl2 with a rounded edge is output.

  On the other hand, the high-pass filter 1H0 outputs a high-frequency signal Sh1 that is a high-frequency signal component of an edge portion that protrudes in the positive amplitude direction, and the high-pass filter 2H0 has a wider width in the positive amplitude direction. The high frequency signal Sh2 of the protruding edge portion is output.

  FIG. 15B is a diagram showing the waveform of each signal when wavelet inverse transformation is performed without performing coring processing.

  In the wavelet inverse transform processing unit 93, processing in accordance with the procedure from the top to the bottom of FIG.

  That is, the synthesis filter 2A0 adds the high-frequency signal Sh2 of the original signal edge portion to the low-frequency signal Sl2 having a sharp edge portion, thereby restoring the slope of the edge and equaling the output signal S11 of the low-pass filter 1L0. The signal is restored.

  Further, by adding the high-frequency signal Sh1 at the original signal edge portion to the signal restored by the synthesis filter 2A0, the slope of the edge is further restored, and a signal equal to the original signal S0 is restored as a restored signal.

  FIG. 15C is a diagram showing the waveform of each signal when wavelet inverse transformation is performed on the signal subjected to coring processing.

  Similarly to FIG. 15B, the wavelet inverse transform processing unit 93 executes processing along the procedure from the top to the bottom of FIG.

  That is, the high-frequency signal Sh2 from the high-pass filter 2H0 is converted into a high-frequency signal Sh2 'whose amplitude value is reduced by the threshold value Th2 for coring processing.

  For this reason, when the high frequency signal Sh2 ′ having a reduced amplitude value is added to the low frequency signal Sl2 by the synthesis filter 2A0, the amplitude value is reduced, so that the one above the edge of the output signal S11 of the low pass filter 1L0 is reduced. A signal having a step in the part is generated.

  Further, the high-frequency signal Sh1 from the high-pass filter 1H0 is also converted into a high-frequency signal Sh1 'whose amplitude value is reduced by the threshold value Th1 for coring processing.

  For this reason, as shown in FIG. 15C, the synthesis filter 1A0 generates a restoration signal having a stepped step at a part of the upper edge of the original signal S0.

  That is, in the case of the configuration shown in FIG. 13, the amount of depression (Th / 4) that is a quarter of the threshold value Th2 as shown in the lowermost part of FIG. 15C with respect to the edge of the original signal S0. , And a dent of (Th2 / 4 + Th1 / 2), which is obtained by adding a quarter of the threshold Th2 and a half of the threshold Th1.

  As a result, the restored signal whose amplitude value is reduced stepwise by (Th2 / 4) and (Th2 / 4 + Th1 / 2) from the flat portion above the edge of the original signal S0 is restored at the upper edge of the restored signal.

  Due to such a stepped step, the edge of the restored signal becomes a rotationally asymmetric edge with respect to the center of the edge, and the edge width of the edge waveform of the original signal is increased due to such a collapse of the edge symmetry. The edge becomes wide.

  As a result, the edge becomes visually blurred, and a phase shift that does not determine the center phase of the edge occurs.

  For this reason, for example, when image processing is performed on each of the RGB signals, the edges are blurred and the phases of the RGB signals are shifted, and a false color contour is formed at the boundary in the image. There was a disadvantage that it was very unsightly.

  The present invention has been made to solve the above problems, and provides an image processing apparatus capable of suppressing edge blur and phase shift due to coring processing while reducing noise by coring processing. Objective.

  The image processing apparatus of the present invention converts a subband image signal for each frequency band into which an image signal is divided into a low frequency band including a low frequency signal component and one or more high frequency bands including a high frequency signal component. A band converting unit; a coring processing unit that performs a coring process for converting a signal component value of a minute amplitude value of the sub-band image signal including the high-frequency signal component into a zero value; and a sub-channel including the low-frequency signal component A subband inverse transform unit that generates an output image signal by combining a band image signal and the coring-processed subband image signal, and the subband inverse transform unit includes the low-frequency signal component One restoration unit that restores one image signal by synthesizing one or more subband image signals subjected to coring processing as a normal phase or a reverse phase according to one combination with a subband image signal; The one or more subband image signals subjected to the coring process are combined with the subband image signal including the low-frequency signal component as the normal phase or the reverse phase according to the other combination different from the one combination. And a restored image synthesizing unit that generates the output image signal by synthesizing the one image signal and the other image signal. .

  With this configuration, the restored image synthesis unit synthesizes one asymmetric edge signal generated by the coring process and the other asymmetric edge signal in the edge signal of the image.

  Here, the one or more subband image signals including the high-frequency signal component converted by the subband conversion unit are classified into a coring target and a coring non-target, and the coring processing unit The coring processing is performed only on the subband image signal including the high frequency signal component to be coring.

  With this configuration, it is possible to select a frequency band for performing coring processing, and it is possible to reduce noise of a specific frequency component.

  In addition, when the absolute value of the supplied signal is equal to or less than a predetermined threshold, the coring processing unit converts the signal into a zero value and outputs the signal, and the absolute value of the supplied signal exceeds the predetermined threshold In some cases, a coring process is performed in which the absolute value of the signal is converted into a value obtained by subtracting the threshold value and output.

  With this configuration, a signal whose absolute value is equal to or less than the threshold value is converted to a zero value, and a signal whose absolute value exceeds the threshold value is converted to a value obtained by subtracting the threshold value from the absolute value, thereby reducing noise.

  Moreover, the said coring process part was set as the structure which performs a coring process based on the predetermined threshold value set to each with respect to the supplied subband image signal containing the said high frequency signal component.

  With this configuration, it is possible to change the amount of noise reduction in the frequency band where the coring process is performed.

  The other restoration unit restores the other image signal which is an image signal delayed by a predetermined pixel from the one image signal, and the restored image synthesis unit converts the one image signal to a predetermined value. The output image signal is generated by synthesizing the image signal delayed by the pixel and the other image signal.

  With this configuration, one image signal and the other image signal can be synthesized with a predetermined phase.

  Further, the other restoration unit is configured to restore the other image signal which is an image signal delayed by one pixel from the one image signal.

  With this configuration, the configuration for delay processing or delay can be reduced.

  Further, the other restoration unit is configured to restore the other image signal which is an image signal delayed by one pixel in the horizontal direction from the one image signal.

  With this configuration, it is possible to reduce the configuration for delay processing or delay, and it is possible to set an edge almost in the middle of the asymmetric blur with respect to the edge signal, and to obtain a more symmetric edge signal.

  The subband converter is configured to convert an image signal into a multiple subband image signal divided in a horizontal frequency band and a vertical frequency band based on wavelet transform.

  With this configuration, noise reduction processing is performed on the subband image signals divided in the horizontal direction and the vertical direction.

  Further, the subband conversion unit has a configuration in which the number of divisions into the vertical frequency band is smaller than the number of divisions into the horizontal frequency band.

  With this configuration, it is possible to reduce the processing or configuration for dividing in the vertical direction while maintaining the noise reduction effect.

  An imaging apparatus according to the present invention includes any one of the image processing apparatuses and an imaging element that images a subject, and an imaging unit that supplies an image signal generated by imaging of the imaging element to the image processing apparatus; And a display unit that displays an image signal image-processed by the image processing apparatus.

  With this configuration, the image displayed on the display unit is an image in which color bleeding or the like is suppressed at an edge portion or the like in the image.

  Furthermore, the image processing method of the present invention converts an image signal into a subband image signal for each frequency band divided into a low frequency band including a low frequency signal component and one or more high frequency bands including a high frequency signal component. Performing a coring process for converting a signal component value of a minute amplitude value of the subband image signal including the high-frequency signal component into a zero value; and subband image signal including the low-frequency signal component A step of restoring one image signal by synthesizing one or more subband image signals subjected to coring processing as a normal phase or a reverse phase according to one combination; and a subband image including the low-frequency signal component One or more subband image signals subjected to the coring process are synthesized into a signal as a normal phase or a reverse phase according to the other combination different from the one combination. And restoring the, it has a configuration that includes a step of generating the output image signal by synthesizing the image signal of the image signal of the one other.

  With this configuration, one asymmetric edge signal generated by the coring process and the other asymmetric edge signal are combined in the edge signal of the image.

  Further, the program of the present invention allows a computer to convert an image signal into a subband image signal for each frequency band divided into a low frequency band including a low frequency signal component and one or more high frequency bands including a high frequency signal component. A step of converting, a step of performing a coring process of converting a signal component value of a minute amplitude value of the subband image signal including the high-frequency signal component into a zero value, and a subband image signal including the low-frequency signal component The step of restoring one image signal by synthesizing one or more subband image signals subjected to coring processing as a normal phase or a reverse phase according to one combination, and subband image signal including the low frequency signal component The other one of the coring-processed one or more subband image signals is synthesized as a normal phase or a reverse phase according to the other combination different from the one combination. Reconstructing an image signal, and has a configuration for executing the step of generating the output image signal by combining the said one of the image signal and the other image signals.

  With this configuration, one asymmetric edge signal generated by the coring process and the other asymmetric edge signal are combined in the edge signal of the image.

  According to the present invention, the low-frequency signal component and the coring-processed high-frequency signal component are synthesized with a combination of different normal phase and negative phase to restore the two systems of image signals. Therefore, it is possible to make a waveform that is rotationally symmetric around the edge of the original signal, against the collapse of edge symmetry caused by coring processing of the restored image signal. Thus, it is possible to suppress edge blurring and phase shift due to coring processing while reducing noise.

  Embodiments of the present invention will be described below with reference to the drawings.

(Embodiment)
FIG. 1 is a block diagram showing a configuration of an image processing apparatus according to an embodiment of the present invention.

  As shown in FIG. 1, the image processing apparatus according to the present embodiment converts an input image signal into a low frequency band including a low frequency signal component and a high frequency signal component including a high frequency signal component of the image signal. A wavelet transform processing unit 1 as a subband transforming unit that transforms into a subband image signal for each frequency band divided by the band, and the highband signal component converted by the wavelet transform processing unit 1 By combining the coring processing unit 2 for performing coring processing for converting the value of the signal component of the minute amplitude value in the component into 0 value, the low frequency signal component and the high frequency signal component subjected to coring processing are output. And a wavelet inverse transform processing unit 3 as a subband inverse transform unit for generating an image signal.

  The image processing apparatus according to the present embodiment performs a process of reducing noise included in an image signal by using wavelet transform, which is one of multi-resolution conversions. In the present embodiment, based on the Harr basis in the wavelet transform, the input image signal is divided into a low frequency band and a high frequency signal component including a low frequency signal component which is a low frequency signal component in a two-dimensional frequency band of a horizontal frequency and a vertical frequency. The case where it isolate | separates into a subband with the high frequency band containing the high frequency signal component which is is shown. In addition, as a number of times of separation into subbands, a case will be described in which separation is performed up to three times of level 3 in the horizontal direction and separation is performed only once of level 1 in the vertical direction.

  FIG. 2 is a diagram showing an example of division for wavelet transform under such conditions.

  FIG. 2A is a diagram showing the division of each sample of the image signal in order to separate the subbands up to level 3 in the horizontal direction and up to level 1 in the vertical direction.

  In FIG. 2 (a), signals S00 to S17 indicate input image signals delayed in units of pixels, respectively, and are delayed one pixel at a time in the horizontal direction from signal S00 to signal S07, and from signal S00 to signal S10. An input image signal delayed by one pixel in the vertical direction is shown.

  FIG. 2 (b) is a diagram showing sub-band division that is separated to level 3 in the horizontal direction and level 1 in the vertical direction.

  In FIG. 2B, numbers indicate wavelet transform levels, “L” indicates a low frequency band, and “H” indicates a high frequency band.

  That is, at level 1, the input image signal has a horizontal low band vertical low band “1LL”, a horizontal high band vertical low band “1HL”, a horizontal low band vertical high band “1LH”, and a horizontal high band vertical high band “1HH”. Are separated into two-dimensional frequency bands.

  Further, the subband image signal of the horizontal low band vertical low band “1LL” is separated at level 2 into a horizontal frequency band of horizontal low band “2L” and horizontal high band “2H”.

  Further, the sub-band image signal of the horizontal low band “2L” is separated at level 3 into horizontal frequency bands of the horizontal low band “3L” and the horizontal high band “3H”.

  FIG. 2C is a diagram showing a procedure for generating a subband image signal for each frequency band divided by the subband of FIG. 2B from each sample shown in FIG. That is, FIG. 2C shows the procedure of wavelet transform in the present embodiment.

  FIG. 3 is a diagram illustrating a detailed configuration of each converting unit of the wavelet transform processing unit 1.

  As shown in FIG. 3, the level 1 conversion unit 14 includes four conversion filters 141. The conversion filter 141 performs an operation based on the Harr base as shown in FIG.

  That is, for example, a signal obtained by adding the respective signals is output to the output LL of the conversion filter 141 to which the signals S00, S01, S10, and S11 are supplied.

  In other words, the output LL of the conversion filter 141 is an output of a digital filter that functions as a two-dimensional low-pass filter in the horizontal direction and the vertical direction.

  On the other hand, outputs HL, LH, and HH of the conversion filter 141 are outputs of a digital filter that functions as a high-pass filter in at least the horizontal direction or the vertical direction in order to output a calculation result including subtraction on the input signal.

  In this way, the conversion filter 141 performs a calculation process as shown in FIG. 4A, thereby functioning as a two-dimensional filter. As shown in FIG. Each conversion signal separated into two-dimensional frequency bands is output, namely “1LL”, horizontal high frequency vertical low frequency “1HL”, horizontal low frequency vertical high frequency “1LH”, and horizontal high frequency vertical high frequency “1HH” Is done.

  The level 2 conversion unit 15 includes two conversion filters 151. The conversion filter 151 performs an operation based on the Harr base as shown in FIG.

  That is, for example, a signal obtained by adding the respective signals is output to the output L of the conversion filter 151 to which the conversion signals 1LL00 and 1LL01 are supplied.

  In other words, the output L of the conversion filter 151 is an output of a digital filter that functions as a one-dimensional low-pass filter in the horizontal direction.

  On the other hand, the output H of the conversion filter 151 is an output of a digital filter that functions as a high-pass filter in the horizontal direction in order to output a calculation result including subtraction on the input signal.

  In this way, the conversion filter 151 functions as a one-dimensional filter by performing the arithmetic processing as shown in FIG. 4B, and as shown in FIG. ”And the horizontal high frequency band“ 2H ”are converted into respective converted signals.

  The level 3 conversion unit 16 includes one conversion filter 151. A signal obtained by adding the respective signals is output to the output L of the conversion filter 151 of the level 3 conversion unit 16.

  The output H of the conversion filter 151 is a signal obtained by subtracting the other signal from one signal.

  In this way, the conversion filter 151 performs a calculation process as shown in FIG. 4B, thereby functioning as a one-dimensional filter. As shown in FIG. The respective conversion signals separated into the horizontal frequency band of “3L” and the horizontal high band “3L” are output.

  As shown in FIG. 1, the wavelet inverse transform processing unit 3 converts the low-frequency transform signal 3L00 and the coring-processed high-frequency transform signal into a normal phase or a reverse phase according to one combination of each high-frequency transform signal, The first restoration unit 31 that restores one image signal by combining the low-frequency conversion signal 3L00 and the coring-processed high-frequency conversion signal is different from the above one combination for each high-frequency conversion signal. The second restoration unit 32 restores the other image signal by combining the normal phase or the reverse phase according to the other combination, and one image signal restored by the first restoration unit 31 and the second restoration. A restored image combining unit 33 that generates the output image signal S ′ by combining the other image signal restored by the unit 32;

  In FIG. 1, the sign “+” shown at the input of each synthesizer is a positive phase, and the sign “−” indicates that both signals are synthesized with a reverse phase obtained by inverting the sign of the signal value. ing. For example, the synthesizer 311 synthesizes the signals 3L00 and 3H00 'in the normal phase, that is, more specifically, synthesizes the signals 3L00 and 3H00'. Further, the synthesis unit 321 performs synthesis by setting the signal 3L00 as the normal phase and the signal 3H00 'as the reverse phase, that is, more specifically, by subtracting the signal 3H00' from the signal 3L00.

  The restored image composition unit 33 delays the image signal S00 ′ restored by the first restoration unit 31 by a predetermined pixel and outputs a delayed image signal S00d ′, and the second restoration unit 32. And a combining unit 332 that combines the image signal S17 ′ restored in step S3 ′ with the image signal S00d ′ from the delay unit 331 and outputs an output image signal S ′.

  In such an image processing apparatus, an input image signal S for reducing noise is supplied to a wavelet transform processing unit 1 as a subband transform unit.

  In the wavelet transform processing unit 1, the supplied input image signal S is supplied to the shift register 11 and the line memory 12.

  The input image signal S supplied to the shift register 11 is shifted by a sampling period while being delayed by a sample unit by the shift register 11, and from the shift register 11, a time-series signal, that is, each of horizontal signals in the horizontal direction. Signals S00, S01, S02, S03, S04, S05, S06, and S07 are supplied to the level 1 converter 14.

  The input image signal S supplied to the line memory 12 is delayed by one horizontal period by the line memory 12, and a signal delayed by one horizontal period is output from the line memory 12.

  Further, this signal is supplied to a shift register 13 having the same function as the shift register 11. As a result, the horizontal signals S10, S11, S12, S13, S14, S15, S16, and S17 are output from the shift register 13 to the level 1 converter 14.

  In this way, the horizontal and vertical image sample signals as shown in FIG. 2A are supplied to the level 1 converter 14 while being shifted in the sampling period.

  In general, since noise has a lower correlation in the vertical direction than in the horizontal direction, noise can be reduced by reducing the number of divisions into vertical frequency bands as compared to the number of divisions into horizontal frequency bands. While maintaining the effect of reduction, the line memory for delaying in the vertical direction can be reduced, and an increase in the amount of hardware can be suppressed.

  The level 1 converter 14 performs level 1 wavelet transform using the signals shown in FIG. 2A supplied from the shift register 11 and the shift register 13.

  As a result, the level 1 conversion unit 14 generates four types of level 1 subband image signals (hereinafter referred to as the level 1 conversion results shown in FIG. 2C) from the two samples in the horizontal direction and the two samples in the vertical direction. If necessary, this is called a converted signal) and output.

  That is, for example, the level 1 conversion unit 14 uses the signals S00, S01, S10, and S11 to convert the converted signal 1LL00 including the horizontal low frequency vertical low frequency signal component and the converted signal including the horizontal high frequency vertical low frequency signal component. 1HL00, a conversion signal 1LH00 including a horizontal low frequency vertical high frequency signal component, and a conversion signal 1HH00 including a horizontal high frequency vertical high frequency signal component are generated.

  Similarly, the level 1 conversion unit 14 generates conversion signals 1LL03, 1HL03, 1LH03, and 1HH03 using, for example, signals S06, S07, S16, and S17.

  The converted signals 1LL00, 1LL01, 1LL02, and 1LL03 including the horizontal low frequency vertical low frequency signal components generated by the level 1 conversion unit 14 are supplied to the level 2 conversion unit 15 that performs level 2 wavelet conversion.

  Also, the converted signal including the high frequency signal component generated by the level 1 converting unit 14 is supplied to the coring processing unit 2 described below.

  As shown in FIG. 2 (c), the level 2 conversion unit 15 generates levels from the converted signals 1LL00, 1LL01, 1LL02, and 1LL03 including the horizontal low frequency vertical low frequency signal components generated by the level 1 conversion unit 14, respectively. 2 conversion signals are generated.

  That is, for example, the level 2 conversion unit 15 uses the conversion signals 1LL00 and 1LL01 to generate a conversion signal 2L00 including a horizontal low-frequency signal component and a conversion signal 2H00 including a horizontal high-frequency signal component, and similarly The conversion signals 2L01 and 2H01 are also generated from the signals 1LL02 and 1LL03.

  The conversion signals 2L00 and 2L01 including the horizontal low-frequency signal component generated by the level 2 conversion unit 15 are supplied to the level 3 conversion unit 16 that performs level 3 wavelet conversion.

  The converted signals 2H00 and 2H01 including the high frequency signal component generated by the level 2 converting unit 15 are supplied to the coring processing unit 2 described below.

  As shown in FIG. 2C, the level 3 conversion unit 16 generates a level 3 conversion signal from the conversion signals 2L00 and 2L01 including the horizontal low-frequency signal component generated by the level 2 conversion unit 15.

  That is, the level 3 conversion unit 16 uses the conversion signals 2L00 and 2L01 to generate a conversion signal 3L00 including a horizontal low-frequency signal component and a conversion signal 3H00 including a horizontal high-frequency signal component.

  The converted signal 3L00 including the horizontal low-frequency signal component generated by the level 3 converting unit 16 is supplied to the wavelet inverse transform processing unit 3 described below.

  The converted signal 3H00 including the high frequency signal component generated by the level 3 converting unit 16 is supplied to the coring processing unit 2 described below.

  As described above, the wavelet transform processing unit 1 includes the low frequency band 3L including the low frequency signal component of the input image signal and the high frequency signal component of the input image signal as illustrated in FIG. It converts into each conversion signal which is a subband image signal for every frequency band divided into one or more high frequency bands 1HL, 1LH, 1HH, 2H, and 3H.

  Next, the conversion signal including the high frequency signal component generated by the level 1 conversion unit 14, the level 2 conversion unit 15, and the level 3 conversion unit 16 is supplied to the coring processing unit 2.

  The coring processing unit 2 performs a coring process for removing the minute amplitude component included in the high frequency signal component for the supplied converted signal including the high frequency signal component.

  Hereinafter, the converted signal 3L00 including only the low-frequency signal component is referred to as a low-frequency converted signal, and other high-frequency signal components such as the converted signals 1HL00, 1LH00, 1HH00, 1HL03, 1LH03, 1HH03, 2H00, 2H01, and 3H00. A conversion signal including the above will be referred to as a high-frequency conversion signal.

  The coring processing unit 2 performs non-linear processing called soft coring as shown in FIG. 5 on each converted signal including the supplied high frequency signal component, that is, the amplitude value of the high frequency converted signal.

  As shown in FIG. 5, when the absolute value of the supplied high-frequency conversion signal is equal to or less than a predetermined threshold value Th (hereinafter referred to as a coring amount as appropriate), the coring processing unit 2 uniformly sets the signal to a zero value. When the absolute value of the converted signal exceeds a predetermined threshold Th, it is converted into a value obtained by subtracting the absolute value of the signal by the threshold Th and output.

  By performing such coring processing in the coring processing unit 2, the signal component of the minute amplitude value included in the supplied high-frequency conversion signal is regarded as noise and converted to 0 value, thereby reducing noise. Will be.

  Each high-frequency transform signal subjected to coring processing by the coring processing unit 2 is supplied to a wavelet inverse transform processing unit 3 described below.

  In the following description, the coring processing unit 2 will be described on the assumption that each high-frequency conversion signal is subjected to coring processing. However, the coring processing may or may not be performed according to each high-frequency conversion signal. It may be a configuration.

  Further, a configuration in which a coring amount corresponding to each high-frequency conversion signal is provided may be used. By adopting such a changeable configuration, it is possible to reduce noise having a specific frequency component or to reduce noise. This makes it possible to perform detailed noise reduction processing, such as changing the amount of noise reduction.

  As described above, the coring processing unit 2 converts the value of the signal component of the minute amplitude value in the high frequency signal component to 0 value with respect to the generated high frequency signal component converted by the wavelet transform processing unit 1. Convert to

  Next, the low-frequency transform signal 3L00 including only the low-frequency signal component generated by the wavelet transform processing unit 1 and the high-frequency transform signal subjected to coring processing by the coring processing unit 2 are supplied to the wavelet inverse transform processing unit 3 Is done.

  In the wavelet inverse transform processing unit 3, the first restoration unit 31 includes a low-frequency transform signal 3L00 including a low-frequency signal component generated by the sub-band division shown in FIG. 2B, and such sub-band division. By combining the high-frequency transform signals including the high-frequency signal components generated by the coring process according to the procedure shown in FIG. 6, the wavelet inverse transform process is executed, and the image signal with reduced noise is restored. To do.

  In the first restoration unit 31, the synthesizing unit 311 adds and synthesizes the low-frequency conversion signal 3L00 and the high-frequency conversion signal 3H00 ′ subjected to the coring process, and further reduces the amplitude of the signal generated by the addition synthesis to one half. By doing so, as shown in FIG. 6, a synthesized signal 2L00 ′ including a low-frequency signal component corresponding to level 2 is generated.

  That is, as shown in FIG. 7A, in the case of horizontal division, when the low-frequency conversion signal L and the high-frequency conversion signal H are added, the signal corresponding to the division position A is restored and low-frequency conversion is performed. When the high frequency conversion signal H is subtracted from the signal L, the signal corresponding to the division position B is restored. Therefore, by adding and synthesizing the low-frequency conversion signal 3L00 and the high-frequency conversion signal 3H00 ', the synthesized signal 2L00' corresponding to the division position 2L00 is restored.

  Next, the synthesizing unit 312 adds and synthesizes the synthesized signal 2L00 ′ generated by the synthesizing unit 311 and the high-frequency conversion signal 2H00 ′ that has been subjected to the coring process, and further reduces the amplitude of the signal generated by the addition and synthesis to two By setting it to 1, as shown in FIG. 6, a synthesized signal 1LL00 ′ including a low frequency signal component corresponding to level 1 is generated.

  Further, the synthesis unit 313 adds and synthesizes the synthesized signal 1LL00 ′ including the low-frequency component generated by the synthesis unit 312 and the coring-processed high-frequency converted signals 1HL00 ′, 1LH00 ′, and 1HH00 ′, and further adds them. By reducing the amplitude of the signal generated by the synthesis to ¼, as shown in FIG. 6, the image signal S00 ′ reduced in noise corresponding to S00 of the image signal is restored.

  That is, as shown in FIG. 7B, in the horizontal and vertical division, when the low-frequency conversion signal LL and the high-frequency conversion signals HL, LH, and HH are added, a signal corresponding to the division position A is obtained. For example, when the high-frequency conversion signals HL and LH are subtracted from the low-frequency conversion signal LL and the high-frequency conversion signal HH is further added, the signal corresponding to the division position D is recovered.

  For this reason, the image signal S00 'corresponding to the signal position S00 is restored as a restoration signal by adding and synthesizing the synthesized signal 1LL00' and the high-frequency conversion signals 1HL00 ', 1LH00', and 1HH00 '.

  As described above, in the high frequency conversion signal, the first restoration unit 31 sets the signal 3H00 ′ as the positive phase, the signal 2H00 ′ as the positive phase, the signal 1HL00 ′ as the positive phase, the signal 1LH00 ′ as the positive phase, and the signal 1HH00. These high-frequency conversion signal and low-frequency conversion signal 3L00 are added in accordance with a combination in which 'is a positive phase.

  In this way, the first restoration unit 31 of the present embodiment synthesizes the converted signals by combining the high-frequency converted signals as one combination, that is, all the positive phases, and one image signal, that is, the image signal S00 ′. To restore.

  The image processing apparatus according to the present embodiment is characterized in that the wavelet inverse transform processing unit 3 further includes a second restoration unit 32 as the other restoration unit.

  That is, in the wavelet inverse transform processing unit 3, the second restoration unit 32 also includes the low-frequency transform signal 3L00 including the low-frequency signal component generated by the subband division shown in FIG. The high-frequency transform signal including the high-frequency signal component generated by the band division and including the coring process is synthesized by the procedure shown in FIG. 6, thereby executing the wavelet inverse transform process and reducing the noise. To restore.

  In the second restoration unit 32, the synthesizing unit 321 subtracts and synthesizes the high-frequency conversion signal 3H00 ′ subjected to the coring process from the low-frequency conversion signal 3L00, and further halves the amplitude of the signal generated by the subtraction synthesis. As a result, a synthesized signal 2L01 ′ including a low-frequency signal component corresponding to level 2 is generated as shown in FIG.

  That is, since the synthesis unit 321 performs subtraction synthesis, as shown in FIG. 7A, the synthesized signal 2L01 'corresponding to the division position 2L01 is restored.

  Next, the synthesizing unit 322 subtracts and synthesizes the high-frequency conversion signal 2H01 ′ subjected to the coring process from the synthesized signal 2L01 ′ generated by the synthesizing unit 321, and further reduces the amplitude of the signal generated by the subtraction synthesis to one half. By doing so, as shown in FIG. 6, a synthesized signal 1LL03 ′ including a low-frequency signal component corresponding to level 1 is generated.

  Further, the synthesizing unit 323 subtracts each of the high-frequency converted signals 1HL03 ′ and 1LH03 ′ subjected to coring processing from the synthesized signal 1LL03 ′ including the low-frequency signal component generated by the synthesizing unit 322, and also outputs the high-frequency converted signal 1HH03. By adding 'and further reducing the amplitude of the signal generated by this arithmetic synthesis to 1/4, as shown in FIG. 6, the noise-reduced image signal S17' corresponding to S17 of the image signal is restored. To do.

  That is, since the synthesis unit 323 performs a calculation corresponding to the calculation (b4) shown in FIG. 7B, the image signal S17 'corresponding to the signal position S17 is restored as a restoration signal.

  As described above, the second restoration unit 32 as the other restoration unit makes the signal 3H00 ′ out of phase, the signal 2H01 ′ out of phase, the signal 1HL03 ′ out of phase, and the signal 1LH03 ′ out of the high frequency conversion signal. These high-frequency conversion signal and low-frequency conversion signal 3L00 are added in accordance with a combination in which the phase is reversed and the signal 1HH03 ′ is in the normal phase.

  In this way, the second restoration unit 32 combines the converted signals with the high-frequency converted signal as a normal phase or a reverse phase according to the other combination, and restores the other image signal, that is, the image signal S17 ′. .

  Further, in the wavelet inverse transform processing unit 3, the restored image synthesis unit 33 further synthesizes the image signal restored by the first restoration unit 31 and the image signal restored by the second restoration unit 32 and outputs them. An image signal S ′ is generated.

  As shown in FIG. 6, the image signal S00 'restored by the first restoration unit 31 and the image signal S17' restored by the second restoration unit 32 have different pixel positions.

  The delay unit 331 is provided to synthesize signals at different pixel positions in this way at a predetermined pixel position.

  The synthesizer 332 synthesizes the image signal S00d 'and the image signal S17' when the image signal S00 'reaches a position corresponding to a predetermined pixel position. The synthesis unit 332 adds the image signal S17 ′ restored by the second restoration unit 32 and the image signal S00d ′ from the delay unit 331, and further halves the amplitude of the signal generated by the addition. Thus, the noise-reduced output image signal S ′ is restored.

  As described above, the image processing apparatus according to the present embodiment has a function of executing the two wavelet inverse transformation processes of the first restoration unit 31 and the second restoration unit 32 in the wavelet inverse transformation processing unit 3. By combining the output signal of two systems by the combining unit 332 as the output image signal S ′, edge blur and phase shift due to coring processing are suppressed while noise is reduced by coring processing. An image processing apparatus is realized.

  Hereinafter, the operation of the image processing apparatus of the present embodiment configured as described above will be described focusing on the operation for the edge of the image, which is a feature of the present invention.

  FIG. 8 illustrates a basic operation in which the first restoration unit 31 and the second restoration unit 32 perform wavelet inverse transformation processing on the horizontal edge signal of the image in the wavelet inverse transformation processing unit 3. It is the wave form diagram shown in order to do.

  FIG. 8A is a diagram illustrating a state of edge repair performed by the synthesis unit 311 of the first restoration unit 31. FIG. 8B is a diagram illustrating a state of edge repair performed by the synthesis unit 321 of the second restoration unit 32.

  First, the synthesis unit 311 of the first restoration unit 31 has an edge portion with respect to the input image signal S from the level 3 conversion unit 16 of the wavelet transform processing unit 1 as shown in FIG. The low-frequency conversion signal 3L00 and the high-frequency conversion signal 3H00 ′ including the high-frequency signal component of the edge portion protruding in the positive amplitude direction are supplied from the coring processing unit 2.

  The synthesizing unit 311 adds the high-frequency conversion signal 3H00 ′ including the high-frequency signal component of the original signal edge portion to the low-frequency conversion signal 3L00 in which the edge portion is rounded, thereby restoring the slope of the edge, and converting the converted signal The output signal 2L00 ′ corresponding to 2L00 is restored.

  At this time, as described above, the converted signal 3H00 including the high frequency signal component of the edge portion is subjected to coring processing by the coring processing unit 2, and thus the high frequency whose amplitude value is decreased by the threshold Th3 of the coring processing. Converted to a converted signal 3H00 ′.

  Therefore, as shown in FIG. 8A, the edge signal synthesized by the synthesis unit 311 is downward from the flat part above the edge by a half of the threshold Th3 (Th3 / 2). The edge signal has a depressed step.

  On the other hand, the low frequency conversion signal 3L00 is supplied from the level 3 conversion unit 16 of the wavelet conversion processing unit 1 and the high frequency conversion signal 3H00 ′ is supplied from the coring processing unit 2 to the synthesis unit 321 of the second restoration unit 32. .

  Here, the synthesizing unit 321 adds and synthesizes the high-frequency conversion signal −3H00 ′, which is the high-frequency conversion signal 3H00 ′ that is inverted in sign, that is, reverse phase to the low-frequency conversion signal 3L00.

  That is, as shown in FIG. 8B, the synthesizer 321 includes the high frequency signal component of the high frequency signal component of the edge portion that protrudes in the negative amplitude direction in the low frequency conversion signal 3L00 with the sharp edge portion. By adding the conversion signal −3H00 ′, the slope of the edge is repaired, and the output signal 2L01 ′ corresponding to the conversion signal 2L01 is restored.

  At this time, as described above, the high-frequency conversion signal 3H00 is converted into the high-frequency conversion signal 3H00 ′ whose amplitude value is reduced by the threshold Th3. The signal waveform decreases by Th3 and protrudes in the negative direction.

  Therefore, as shown in FIG. 8B, the edge signal synthesized by the synthesis unit 321 is upward from the flat part below the edge by a half amount of the threshold Th3 (Th3 / 2). The edge signal has a step that is recessed in the area.

  FIG. 9 is a diagram for explaining the operation of the restored image composition unit 33 in the wavelet inverse transform processing unit 3. Here, the coring process is performed with the coring amount set to the threshold value Th3 for the level 3 high frequency conversion signal 3H00, and the coring value is set to the threshold value Th2 with respect to the level 2 high frequency conversion signal 2H00. An example in which the coring process is performed and the coring amount is set to the threshold value Th1 for the level 1 high-frequency conversion signals 1HL00 and 1HL03 will be described.

  9, FIG. 9A shows an image signal S00 ′ output from the first restoration unit 31, FIG. 9B shows an image signal S17 ′ output from the second restoration unit 32, and FIG. ) Shows the waveform and timing of the signal S00d ′ output from the delay unit 331, and FIG. 9D shows the waveform and timing of the output image signal S ′ output from the synthesis unit 332.

  Hereinafter, with reference to FIG. 9, the operation for suppressing the influence of the edge difference as shown in FIG. 8 by the image processing apparatus of the present embodiment will be mainly described.

  As shown in FIG. 9A, the edge waveform of the image signal S00 ′ reconstructed by the first reconstructing unit 31 is composed of a plurality of steps by synthesizing with each converted signal including a high-frequency component subjected to coring processing. Edge waveform including That is, since coring processing is performed in three stages from level 3 to level 1, the edge waveform of the image signal S00 ′ starts from the flat portion above the edge from the high frequency conversion signal 3H00 ′, the high frequency conversion signal 2H00 ′, and the high frequency conversion. It has a stepped waveform that is depressed downward by a reduction amount due to the signal 1HL00 ′ or the like.

  As the amount of decrease from the flat portion above the edge, first, in the region affected only by the threshold value Th3 of the coring process for the high-frequency conversion signal 3H00, the calculation (a1) in FIG. As a result, the amount of decrease corresponding to the threshold value Th3 is halved, resulting in (Th3 / 2) as shown in FIG. 8, and further divided by the calculation (a1) in FIG. As a result, (Th3 / 4) is obtained, and as a result of being further reduced to a quarter by the calculation (b1) of FIG. Becomes (Th3 / 16).

  Further, as a reduction amount in the region affected by the threshold Th3 for the high-frequency conversion signal 3H00 and the threshold Th2 for the high-frequency conversion signal 2H00, in addition to the reduction amount (Th3 / 16) in the region affected only by the threshold Th3, synthesis is performed. As a result of the reduction amount corresponding to the threshold value Th2 being halved by the calculation (a1) in FIG. 7A in the unit 312, the result is (Th2 / 2), and the calculation (b) in FIG. As a result of being further reduced to a quarter by b1), (Th2 / 8) is obtained. As a result, as shown in FIG. 9, the reduction amount of this region is (Th3 / 16 + Th2 / 8).

  Further, as a reduction amount in a region affected by the threshold Th3 for the high-frequency conversion signal 3H00, the threshold Th2 for the high-frequency conversion signal 2H00, and the threshold Th1 for the high-frequency conversion signal 1HL00, the threshold Th3 for the high-frequency conversion signal 3H00 and the high The reduction amount corresponding to the threshold value Th1 is calculated as 4% by the calculation (b1) of FIG. 7B in the combining unit 313 in the reduction amount (Th3 / 16 + Th2 / 8) in the region where the threshold value Th2 affects the area conversion signal 2H00. As a result, (Th1 / 4), which is a result of 1, is added. As a result, as shown in FIG. 9, the reduction amount of this region is (Th3 / 16 + Th2 / 8 + Th1 / 4).

  For this reason, as described above, the edge waveform of the image signal S00 'is an edge waveform whose edge width is wider than the edge waveform of the original signal, and is a visually blurred edge. Further, the edge waveform of the image signal S00 'spreads in the time direction from the edge waveform of the original signal, so that the edge shift occurs and the center phase of the edge is not determined.

  Further, as shown in FIG. 9B, the edge waveform of the image signal S17 ′ reconstructed by the second reconstructing unit 32 is also combined with each converted signal including a high-frequency component subjected to coring processing. The edge waveform includes the step. That is, since coring processing is performed in three stages from level 3 to level 1, the edge waveform of the image signal S17 ′ starts from the flat portion below the edge from the high frequency conversion signal 3H00 ′, the high frequency conversion signal 2H01 ′, and the high frequency It has a stepped waveform that is depressed upward by a reduction amount due to the conversion signal 1HL03 ′ or the like. The amount of decrease from the flat portion below the edge is the same as in the case of the first restoration portion 31, and (Th3 / 16), (Th3 / 16 + Th2 / 8), and (Th3 / 16 + Th2 / 8 + Th1). / 4), which is a three-step level difference.

  For this reason, the edge waveform of the image signal S17 'is also an edge waveform whose edge width is wider than the edge waveform of the original signal, and is a visually blurred edge. Further, since the edge waveform of the image signal S17 'spreads in the direction opposite to the time from the edge waveform of the original signal, the edge shift occurs and the edge center phase is not determined.

  Further, since the first restoration unit 31 restores the image signal S00 ′ corresponding to the pixel position S00, and the second restoration unit 32 simultaneously restores the image signal S17 ′ corresponding to the pixel position S17, FIG. ) And the pixel position of the restored signal are shifted as shown in FIG. Therefore, the delay unit 331 delays the image signal S00 'until the image signal S00' reaches a predetermined pixel position.

  FIG. 9C shows the timing of the image signal S00d 'obtained by delaying the image signal S00' by a predetermined delay amount by the delay unit 331. That is, as shown in FIGS. 9B and 9C, a predetermined phase is set such that the phase of the original signal in the image signal S17 ′ is the same as the phase of the original signal in the delayed image signal S00d ′. The delay unit 331 delays the image signal S00 ′ by the delay amount.

  The synthesizing unit 332 averages the image signal S17 ′ restored by the second restoring unit 32 as shown in FIG. 9B and the image signal S00d ′ from the delay unit 331 as shown in FIG. 9C. Perform value calculation. FIG. 9D shows the edge waveform of the output image signal S ′ output from the combining unit 332 as a result of the average value calculation of the image signal S17 ′ and the image signal S00d ′.

  As shown in FIG. 9D, first, the waveform of the signal synthesized by the synthesis unit 332 becomes a waveform that is rotationally symmetric about the edge of the original signal. That is, since the synthesizing unit 332 synthesizes each waveform portion in which the slope of the original signal is restored without being affected by the coring process, the vicinity of the center of the output image signal S ′ obtained by this synthesis is As a result, the center phase of the edge is determined as shown in FIG. 9D.

  Therefore, according to the image processing apparatus of the present embodiment, noise can be reduced by the coring processing in the coring processing unit 2, and a phase shift due to the coring processing can also be suppressed. For this reason, for example, in the case where image processing is performed on each of the RGB signals and displayed, the phases of the RGB signals can be matched, and generation of false color contours and the like can be prevented.

  Further, since the synthesizing unit 332 takes the average value of the image signal S17 ′ and the image signal S00d ′, the range of the stepped waveform is widened, but the amount of reduction from the flat portion of the edge to the step portion is the image The amount is reduced by half compared to the decrease amount of the signal S17 ′ and the image signal S00d ′.

  That is, in the case of the image signal S00 ′ shown in FIG. 9A, the amount of decrease from the flat portion on the upper edge is affected only by the threshold Th3 (Th3 / 16), and the threshold Th3 and the threshold Th2 are affected (Th3). / 16 + Th2 / 8) and the threshold Th3 to the threshold Th1 are affected (Th3 / 16 + Th2 / 8 + Th1 / 4).

  On the other hand, as shown in FIG. 9D, in the case of the output image signal S ′, the threshold Th3 and the threshold Th2 are affected only by the threshold Th3 (Th3 / 32). (Th3 / 32 + Th2 / 16), and the threshold Th3 to the threshold Th1 are affected (Th3 / 32 + Th2 / 16 + Th1 / 8), which is half the reduction amount of the image signal S17 ′ and the image signal S00d ′. Become. As described above, although the stepped waveform range is widened, the intensity of image blur can be halved, so that the edge blur due to the coring process can be made inconspicuous.

  As described above, according to the image processing apparatus of the present embodiment, it is possible to suppress edge blur and phase shift due to coring processing while reducing noise by coring processing.

  FIG. 10 is a block diagram showing another aspect of the image processing apparatus of the present embodiment. The image processing apparatus shown in FIG. 10 is different from the image processing apparatus shown in FIG. 1 in that the first restoration unit 41 and the second restoration unit 42 of the wavelet inverse transformation processing unit 4 are different from those in FIG. It is the structure which restores.

  In FIG. 10, a wavelet transform processing unit 1 and a coring processing unit 2 have the same configuration as the image processing apparatus in FIG. FIG. 11 is a diagram illustrating a procedure for restoring an output image signal with reduced noise from the generated subband image signal in the configuration of FIG. 10.

  Hereinafter, the configuration and operation of the image processing apparatus illustrated in FIG. 10 will be described. Similarly to the configuration described in FIG. 1, the wavelet inverse transform processing unit 4 shown in FIG. 10 converts the low-frequency transformed signal 3L00 and the coring-processed high-frequency transformed signal into one high-frequency transformed signal according to one combination. A first restoration unit 41 that restores one image signal by combining the normal phase or the reverse phase and combining the low-frequency conversion signal 3L00 and the coring-processed high-frequency conversion signal with respect to each high-frequency conversion signal One image restored by the first restoration unit 41 and the second restoration unit 42 which restores the other image signal by combining the normal phase or the reverse phase according to the other combination different from the one combination. A restored image synthesis unit 43 that generates an output image signal S ′ by synthesizing the signal and the other image signal restored by the second restoration unit 42 is provided.

  In the first restoration unit 41, the synthesis unit 411 adds and synthesizes the low-frequency conversion signal 3L00 and the high-frequency conversion signal 3H00 ′ that has been subjected to the coring process, and further halves the amplitude of the signal generated by the addition synthesis. By doing so, as shown in FIG. 11, a synthesized signal 2L00 ′ including a low-frequency signal component corresponding to level 2 is generated.

  Next, the synthesizing unit 412 subtracts and synthesizes the high-frequency conversion signal 2H00 ′ that has been subjected to the coring process from the synthesized signal 2L00 ′ generated by the synthesizing unit 411, and further reduces the amplitude of the signal generated by the subtractive synthesis to one half. By doing so, as shown in FIG. 11, a synthesized signal 1LL01 ′ including a low-frequency signal component corresponding to level 1 is generated.

  Further, the synthesizer 413 subtracts the high-frequency converted signals 1HL01 ′ and 1HH01 ′ subjected to coring processing from the synthesized signal 1LL01 ′ including the low-frequency component generated by the synthesizer 412 and also outputs the high-frequency converted signal 1LH01 ′. By adding and further reducing the amplitude of the signal generated by this arithmetic synthesis to 1/4, as shown in FIG. 11, the noise-reduced image signal S03 ′ corresponding to S03 of the image signal is restored.

  On the other hand, in the second restoration unit 42, the synthesizing unit 421 subtracts and combines the high-frequency conversion signal 3H00 ′ that has been coring-processed from the low-frequency conversion signal 3L00, and further reduces the amplitude of the signal generated by subtraction synthesis to 2 minutes. By setting it to 1, as shown in FIG. 11, a synthesized signal 2L01 ′ including a low frequency signal component corresponding to level 2 is generated.

  Next, the synthesizing unit 422 adds and synthesizes the synthesized signal 2L01 ′ generated by the synthesizing unit 421 and the high-frequency conversion signal 2H01 ′ that has been subjected to the coring process, and further reduces the amplitude of the signal generated by the adding and synthesizing by two. By setting it to 1, as shown in FIG. 11, a synthesized signal 1LL02 ′ including a low-frequency signal component corresponding to level 1 is generated.

  Further, the synthesis unit 423 adds the synthesized signal 1LL02 ′ including the low-frequency signal component generated by the synthesis unit 422 and the high-frequency converted signals 1HL02 ′, 1LH02 ′, and 1HH02 ′ that have been subjected to the coring process, and further performs addition synthesis By reducing the amplitude of the signal generated by ¼, as shown in FIG. 11, the noise-reduced image signal S04 ′ corresponding to S04 of the image signal is restored.

  The restored image composition unit 43 delays the image signal S03 ′ restored by the first restoration unit 41 by a predetermined pixel and outputs a delayed image signal S03d ′, and the second restoration unit 42. In this configuration, the image signal S04 ′ restored in step S4 ′ and the image signal S03d ′ from the delay unit 431 are averaged and combined to output an output image signal S ′.

  In the configuration of FIG. 10, as shown in FIG. 11, the image signal S03 'and the image signal S04' whose pixel positions are adjacent are restored simultaneously. That is, the second restoration unit 42 is configured to restore the other image signal S04 ′, which is an image signal delayed by one pixel in the horizontal direction from the one image signal S03 ′ restored by the first restoration unit 41. is there. With such a configuration, the pixel positions of the restored image signal are adjacent to each other, so that the hardware amount of the delay unit 431 can be reduced, and the edge signal is asymmetric by the synthesis in the synthesis unit 432. The edge can be set almost in the middle of the edge, and an output image signal S ′ having a more symmetrical edge signal can be obtained.

  As described above, according to the image processing apparatus of the present embodiment, in the inverse wavelet transform processing unit, the restored image synthesis unit uses the one image signal restored by the first restoration unit and the second restoration unit. The restored other image signal is synthesized by averaging, and the synthesized signal is output as an output image signal.

  That is, the restored image synthesis unit synthesizes one asymmetric edge signal generated by the coring process and the other asymmetric edge signal with a predetermined phase in the edge signal of the image.

  For this reason, the edge waveform in the output image signal can suppress the influence of the step caused by the coring process, and can be output as a rotationally symmetric edge waveform.

  Therefore, according to the image processing apparatus of the present embodiment, it is possible to provide an image processing apparatus in which edge blurring and phase shift due to coring processing are suppressed while noise is reduced by coring processing.

  FIG. 12 is a block diagram illustrating a configuration example of an apparatus using the image processing apparatus according to the present embodiment. Here, an example of an imaging apparatus that electronically images a subject such as an electronic camera is given.

  As illustrated in FIG. 12, the imaging apparatus performs image processing on an imaging unit 60 that is an optical system for imaging a subject and an imaging signal from the imaging unit 60, and outputs a signal that has been subjected to image processing A signal processing unit 50 that outputs an image signal and a display unit 70 that displays the output image signal are provided.

  The imaging unit 60 includes a lens unit 61 that images a subject, and an imaging element 62 that captures an image formed by the lens unit 61. The image sensor 62 converts the image formed by the lens unit 61 into an electrical signal and supplies it to the signal processing unit 50 as an imaging signal.

  In the signal processing unit 50, first, an imaging signal from the imaging device 62 is supplied to an analog front end (hereinafter referred to as AFE) 51 of the signal processing unit 50. The AFE 51 performs an amplification process on the supplied imaging signal and supplies the amplified imaging signal to an AD (Analog to Digital) converter 52.

  The AD converter 52 converts an imaging signal, which is an analog signal, into a digital signal, and the imaging data converted into the digital signal is supplied as an image signal to the image processing device 55 of the present embodiment.

  The image processing device 55 reduces the noise of the image signal, and the noise-reduced image signal is supplied to the DA (Digital to Analog) converter 53 and the recording unit 54. The DA converter 53 converts an image signal, which is a digital signal, into an analog signal, and the analog signal is supplied to the display unit 70.

  As a result, a noise-reduced subject image is displayed on the display unit 70. Further, the recording unit 54 includes, for example, a recording medium that records image data, and an image of a subject with reduced noise is recorded on the recording medium of the recording unit 54.

  The control unit 59 controls the signal processing unit 50 as a whole. In particular, the image processing apparatus 55 according to the present embodiment is characterized in that edge phase shift is suppressed as described above. For this reason, in the present imaging device, an image displayed on the display unit 70 is an image in which color bleeding or the like is suppressed at an edge portion or the like in the image.

  In the above description, the number of times that the input image signal is separated into subbands of the low frequency band and the high frequency band in the two-dimensional frequency band of the horizontal and vertical frequencies based on the Harr basis in the wavelet transform, and the number of times of separation into subbands. In the above description, the level direction is separated up to level 3 and only the level 1 is separated in the vertical direction. However, the present invention is not limited to these. The present invention can also be applied to an image processing apparatus using transformation, and can also be applied to an image processing apparatus using wavelet transformation and inverse transformation based on a base other than the Harr basis. Further, the separation into subbands is not limited to the two-dimensional frequency band, and the level to be separated may be frequency separated to an arbitrary level.

  In addition, although an example has been described in which the coring processing unit performs coring processing on each high-frequency conversion signal, the configuration is such that coring processing is selectively performed according to each high-frequency conversion signal. May be.

  That is, each high-frequency signal component converted by the wavelet transform output unit is classified into a coring target and a coring non-target, and the coring processing unit performs a coring process on the coring target high-frequency signal component. In the wavelet inverse transform processing unit, one restoration unit converts the high-frequency signal component of the low-frequency signal component, the high-frequency signal component subjected to coring processing, and the high-frequency signal component that is not subject to coring, respectively, The high-frequency signal component is set to the normal phase or the reverse phase according to one combination, and one image signal is restored by synthesizing, and the other restoration unit includes a low-frequency signal component and a coring-processed high-frequency signal component Each high-frequency signal component with a non-coring high-frequency signal component is positively phased according to the other combination different from one combination for each high-frequency signal component or Phase cities, may be configured so as to restore the other image signal by synthesizing.

  In addition, the coring processing unit may be configured to perform coring processing based on a predetermined threshold set for each of the supplied high frequency signal components.

  The image processing apparatus according to the present embodiment has been described using functional blocks as shown in FIGS. 1 and 10. For example, a wavelet transform, an inverse transform, and a coring process are performed using a microcomputer. May be performed by the procedure as described above.

  That is, the image signal is converted into a subband image signal for each frequency band divided by a low frequency band including a low frequency signal component of the image signal and one or more high frequency bands including a high frequency signal component. A step of performing a coring process on the converted high-frequency signal component, converting the value of the signal component of the minute amplitude value in the high-frequency signal component to a zero value, and coring the low-frequency signal component Each of the high-frequency signal components is converted into a normal phase or a reverse phase according to one of the combinations for each high-frequency signal component, and one image signal is restored by synthesizing the high-frequency signal component; A step of restoring the other image signal by combining the high-frequency signal component into a normal phase or a reverse phase according to the other combination different from the one combination for each high-frequency signal component; A program of an image processing method including a step of generating an output image signal by combining one image signal restored by the restoration unit and the other image signal restored by the other restoration unit. The output image signal may be generated by reading the program recorded in the memory and executing the program according to each step of the program.

  As described above, the image processing apparatus according to the present invention has the effect of suppressing edge blurring and phase shift due to coring processing while reducing noise by coring processing, and the camera of the imaging apparatus. And an information terminal device such as a mobile phone having an imaging function, a television having an image display function, a DVD (Digital Versatile Disk) recorder having an image recording function, and the like.

1 is a block diagram of an image processing apparatus according to an embodiment of the present invention. (A) The figure which showed the division | segmentation of each sample of the image signal of the image processing apparatus in embodiment of this invention (b) The figure which showed the division | segmentation of the subband of the image processing apparatus in embodiment of this invention (c) The figure which showed the procedure of the wavelet transformation of the image processing apparatus in embodiment of this invention The figure which showed the detailed structure of each conversion part of the wavelet transformation process part of the image processing apparatus in embodiment of this invention (A) The figure which shows the calculation process of the conversion filter of the level 1 conversion part of the image processing apparatus in embodiment of this invention (b) The level 2 conversion part and level 3 conversion part of the image processing apparatus in embodiment of this invention Showing the calculation processing of the conversion filter The figure which shows the coring process of the image processing apparatus in embodiment of this invention The figure which showed the procedure of the wavelet inverse transformation of the image processing apparatus in embodiment of this invention (A) The figure which shows the arithmetic expression of the wavelet transformation at the time of the horizontal division of the image processing apparatus in embodiment of this invention, and the decompression | restoration position of a decompression | restoration signal (b) Horizontal and the image processing apparatus in embodiment of this invention The figure which shows the calculation formula of wavelet inverse transform at the time of vertical division, and the restoration position of restoration signal (A) Waveform diagram showing a state of edge repair performed by the synthesis unit of the first restoration unit of the image processing apparatus according to the embodiment of the present invention. (B) Second waveform of the image processing apparatus according to the embodiment of the present invention. Waveform diagram showing the state of edge repair performed by the synthesis unit of the restoration unit (A) The figure which shows the waveform and timing of an image signal output from the 1st decompression | restoration part of the image processing apparatus in embodiment of this invention (b) The 2nd decompression | restoration of the image processing apparatus in embodiment of this invention (C) The figure which shows the waveform and timing of the signal output from the delay part of the image processing apparatus in embodiment of this invention (d) The figure which shows the waveform and timing of the image signal output from a part Showing waveform and timing of output image signal output from combining unit of image processing apparatus in form The block diagram of the other aspect of the image processing apparatus in embodiment of this invention The figure which showed the procedure of the wavelet inverse transformation of the other aspect of the image processing apparatus in embodiment of this invention 1 is a block diagram of an imaging apparatus using an image processing apparatus according to an embodiment of the present invention. Block diagram of a conventional image processing apparatus (A) The figure which shows the example of a sample of the input signal of the conventional image processing apparatus (b) The figure which shows the example of the wavelet transformation of the conventional image processing apparatus (c) The figure which shows the example of the wavelet inverse transformation of the conventional image processing apparatus (A) Waveform diagram of the original signal near the edge in the conventional image processing apparatus and the output signal from each filter of the wavelet transform processing unit (b) Wavelet inverse transform is performed without performing coring processing in the conventional image processing apparatus (C) Waveform diagram of each signal when wavelet inverse transformation is performed on a signal subjected to coring processing in a conventional image processing apparatus

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Wavelet conversion process part 11 Shift register 12 Line memory 13 Shift register 14 Level 1 conversion part 141 Conversion filter 15 Level 2 conversion part 151 Conversion filter 16 Level 3 conversion part 2 Coring process part 3 Wavelet inverse conversion process part 31 1st Restoring unit 311, 312, 313 Combining unit 32 Second restoring unit 321, 322, 323 Combining unit 33 Restoring image combining unit 331 Delay unit 332 Combining unit 4 Wavelet inverse transform processing unit 41 First restoring unit 411, 412, 413 Combining unit 42 Second restoring unit 421, 422, 423 Combining unit 43 Restoring image combining unit 431 Delay unit 432 Combining unit 50 Signal processing unit 51 AFE
52 AD converter 53 DA converter 54 Recording unit 55 Image processing device 59 Control unit 60 Imaging unit 61 Lens unit 62 Imaging element 70 Display unit 91 Wavelet conversion processing unit 92 Coring processing unit 93 Wavelet inverse conversion processing unit 911 Delay element 912 Adder 913 Subtractor 921 Coring circuit 922 Coring circuit 931 Adder 932 Divider

Claims (12)

A subband conversion unit that converts an image signal into a subband image signal for each frequency band divided by a low frequency band including a low frequency signal component and one or more high frequency bands including a high frequency signal component; A coring processing unit for performing a coring process for converting a signal component value of a minute amplitude value of a sub-band image signal including a signal component into a zero value, a sub-band image signal including the low-frequency signal component, and the coring process A subband inverse transform unit that generates an output image signal by combining the generated subband image signal, and the subband inverse transform unit adds the core to the subband image signal including the low-frequency signal component. One restoration unit that restores one image signal by combining one or more ring-processed subband image signals as a positive phase or a reverse phase according to one combination, and the low-frequency signal component The other image signal is restored by synthesizing one or more subband image signals subjected to the coring process into a normal image or a reverse phase according to the other combination different from the one combination. An image processing apparatus comprising: a restoration unit; and a restoration image synthesis unit that generates the output image signal by synthesizing the one image signal and the other image signal. The one or more subband image signals including the high-frequency signal component converted by the subband conversion unit are classified into a coring target and a coring non-target, and the coring processing unit is the coring target The image processing apparatus according to claim 1, wherein coring processing is performed only on a subband image signal including a high-frequency signal component. When the absolute value of the supplied signal is equal to or less than a predetermined threshold, the coring processing unit converts the signal into a zero value and outputs it.When the absolute value of the supplied signal exceeds the predetermined threshold, 3. The image processing apparatus according to claim 1, wherein coring processing is performed for converting and outputting the absolute value of the signal to a value obtained by subtracting the absolute value by a threshold value. The coring processing unit performs coring processing based on a predetermined threshold set for each of the supplied subband image signals including the high-frequency signal component. The image processing apparatus according to 3. The other restoration unit restores the other image signal which is an image signal delayed by a predetermined pixel from the one image signal, and the restored image composition unit converts the one image signal by a predetermined pixel. The image processing apparatus according to claim 1, wherein the output image signal is generated by combining the delayed image signal and the other image signal. The image processing apparatus according to claim 5, wherein the other restoration unit restores the other image signal that is an image signal delayed by one pixel from the one image signal. The image processing apparatus according to claim 5, wherein the other restoration unit restores the other image signal which is an image signal delayed by one pixel in the horizontal direction from the one image signal. 8. The subband conversion unit according to claim 1, wherein the subband conversion unit converts the image signal into a multiple subband image signal divided in a horizontal frequency band and a vertical frequency band based on wavelet conversion. The image processing apparatus according to item 1. The image processing apparatus according to claim 8, wherein the subband conversion unit has a smaller number of divisions into vertical frequency bands than a number of divisions into horizontal frequency bands. An image processing apparatus according to any one of claims 1 to 9, and an imaging unit that includes an imaging element that images a subject, and that supplies an image signal generated by imaging of the imaging element to the image processing apparatus. An image pickup apparatus comprising: a display unit that displays an image signal image-processed by the image processing apparatus. Converting the image signal into a subband image signal for each frequency band divided by a low frequency band including a low frequency signal component and one or more high frequency bands including a high frequency signal component; and A step of performing coring processing for converting a signal component value of a minute amplitude value of a subband image signal including the signal into a zero value, and at least one of the subband image signals including the low frequency signal component subjected to the coring processing. A sub-band image signal of one of the image signals is synthesized by combining the sub-band image signal as a positive phase or a negative phase according to one combination, and the coring process is applied to the sub-band image signal including the low-frequency signal component. Reconstructing the other image signal by synthesizing two or more subband image signals as normal phase or reverse phase according to the other combination different from the one combination; An image processing method characterized by comprising the step of generating the output image signal by synthesizing the image signal and the other image signals. Converting the image signal into a sub-band image signal for each frequency band divided into a low-frequency band including a low-frequency signal component and one or more high-frequency bands including a high-frequency signal component; Performing a coring process for converting a signal component value of a minute amplitude value of a subband image signal including a component into a zero value, and applying the coring process to the subband image signal including the low frequency signal component A step of restoring one image signal by synthesizing the above subband image signals as a positive phase or a reverse phase according to one combination, and the subband image signal including the low-frequency signal component is subjected to the coring process. Reconstructing the other image signal by synthesizing two or more subband image signals as a positive phase or a reverse phase according to another combination different from the one combination, Program for executing the step of generating the output image signal by combining the serial one image signal and the said other image signals.

Images