JP4194472B2 - Image processing method and apparatus - Google Patents

Description

  The present invention relates to digital signal conversion, and more particularly, to image processing for performing discrete wavelet transformation (DWT) on an image.

  In recent years, image coding called JPEG2000 using DWT has been standardized by ISO / IEC (ISO / IEC-15444). JPEG2000 is an encoding method in which an input two-dimensional image is subjected to two-dimensional DWT transformation and then quantized and compressed by arithmetic coding. In the following description, “wavelet transform” may be used to mean “discrete wavelet transform”.

  In JPEG2000, one image is divided into divided areas called tiles, and encoding processing is performed for each tile. The entire image can be encoded as a single tile.

  DWT is also performed in units of tiles. The DWT used in JPEG2000 employs a lifting operation. Lifting calculation is a process of filtering and sub-sampling at the same time by repeating the calculation of adding data obtained by multiplying the sum of one or more existing data before and after a certain data in one-dimensional data. It is a calculation (for example, see Patent Document 1 for lifting calculation in JPEG2000). The lifting operation (forward conversion) of the irreversible filter defined in the JPEG2000 standard is expressed by the following recurrence formula.

However,
α = −1.586 134 342 059 924
β = −0.052 980 118 572 961
γ = 0.882 911 075 530 934
δ = 0.443 506 852 043 971
K = 1.230 174 104 914 001
n is an integer. ... (Formula 1)

  As described above, in order to perform an operation on certain data in the lifting operation, data existing before and after the data is necessary, but the pixel necessary for the operation does not actually exist for the data at the end of the image. Therefore, the JPEG2000 lifting calculation makes up the virtual pixel column by folding the pixel column at the edge of the image. The number of pixel rows to be supplemented is 3 pixels when the coordinates of the end pixels are odd, and 4 pixels when the coordinates are even.

  This pixel folding will be described with reference to FIG. In FIG. 11, when alphabets A to F are actual pixel columns and the coordinates of the pixels are 0 to 5 in order from the pixels A to F, the four pixels B, C, D, and E are in reverse order for the pixel A having an even address. Thus, on the left side of the pixel A, with respect to the pixel F having an odd address, a virtual pixel column in which the column is folded is supplemented to the end of the actual pixel column.

  As a result of the lifting operation, a low frequency component appears when the pixel coordinates are even (Y (2n)), and a high frequency component appears when the pixel coordinates are odd (Y (2n + 1)). Accordingly, the calculation of step 2, 4, 6 in Equation 1 can be regarded as low-pass filter processing, and the calculation of step 1, 3, 5 can be regarded as high-pass filter processing. The DWT repeatedly performs the lifting operation shown in the recurrence formula of Formula 1 in a one-dimensional direction, and divides the image data into a low frequency component and a high frequency component. Further, this calculation is sequentially applied to each pixel line in the horizontal direction from the left end to the right end and the vertical direction from the upper end to the lower end, thereby dividing the image data into two-dimensional subbands.

  FIG. 12 shows a state in which an image of 4 pixels in the vertical direction and 4 pixels in the horizontal direction is divided into subbands by the wavelet transform. L and H in the figure represent frequency components included in the conversion coefficient, L represents a low frequency component, and H represents a high frequency component. The numbers shown in parentheses represent the coordinates of the pixels in the original image, the left-hand value is the vertical coordinate from the upper end to the lower end, and the right-hand value is the horizontal coordinate from the left end to the right end. Represents.

  In the two-dimensional DWT, first, a one-dimensional lifting operation is performed from the left end to the right end for each vertical pixel line from the upper end to the lower end. FIG. 12A shows the result of performing a one-dimensional lifting operation for each vertical pixel line. After the lifting calculation, rearrangement is performed to combine the calculation results for each of the L and H components. The result of rearranging the coefficients is shown in FIG. Subsequently, a one-dimensional lifting operation is performed from the upper end to the lower end for each horizontal pixel line from the left end to the right end. The result is shown in FIG. Subsequently, similar to the calculation for each vertical pixel line, the calculation result is rearranged for each of the L and H components. The result is shown in FIG. As a result, the coefficients generated by the wavelet transform are grouped into subbands as shown in FIG.

  FIG. 13 shows where the subband coefficients generated by performing wavelet transform on an image of four vertical pixels and four horizontal pixels are located in the original image. FIG. 13A shows the distribution of the coefficients generated by the one-dimensional DWT for each vertical pixel line from the upper end to the lower end in the positional relationship of the original pixels, and FIG. 13B shows the two-dimensional DWT. The distribution of the coefficients generated by the above is shown by the positional relationship of the original pixels. FIG. 13 shows that the frequency component of the coefficient generated by the wavelet transform differs depending on whether the pixel line address is an odd number or an even number.

  As described above, since wavelet transform is applied to each tile in JPEG 2000, the distribution of coefficients when the wavelet transform is performed on the tile will be described here. FIG. 14 shows the distribution of coefficients generated by wavelet transform in terms of the original pixel positional relationship when an image of 4 pixels in the vertical direction and 8 pixels in the horizontal direction is divided into two tiles of 4 pixels in the vertical and horizontal directions. . FIG. 14A shows a distribution of coefficients generated by a two-dimensional DWT. FIG. 14B shows only the distribution of frequency components in the same vertical direction as the direction of the Tile division boundary among the coefficients shown in FIG. From FIG. 14B, it can be seen that when Tile division is performed, coefficients of the same frequency component are arranged on the division boundary surface with respect to the frequency component in the same vertical direction as the direction of the Tile division boundary.

JP 2002-152045 A

  Normally, in the lossy compression quantization of JPEG 2000, a coefficient including a high frequency component is quantized in a larger quantization step than a coefficient including a low frequency component. This is to efficiently encode using the human visual characteristic that the visibility of the high frequency component is inferior to the visibility of the low frequency component. As a result, the quantization error of the coefficient including the high frequency component becomes larger and deteriorates than the coefficient including the low frequency component. As described above, since the conversion coefficient of the two-dimensional DWT has a configuration in which a high frequency component and a low frequency component are arranged in accordance with the address of the pixel line, a pixel having a large deterioration and a small pixel in accordance with the address of the pixel line. Will be lined up.

  As described with reference to FIG. 12, in the JPEG2000 lifting calculation, the pixel column is folded at the end of the image to compensate for the virtual pixel column. For this reason, when irreversible conversion is performed, image quality degradation related to the virtual pixel column affects the generation of the actual pixel column during inverse wavelet transform (decoding), and the image quality of the actual pixel column deteriorates. End up.

  Here, the image quality degradation accompanying the interpolation of the virtual pixel column is most significant in the endmost pixel most affected by the virtual pixel column. In the lifting calculation, pixels that need to be interpolated are interpolated using the peripheral pixels, so that the low-frequency component storage is excellent, but the high-frequency component storage is not excellent. For this reason, image quality degradation due to interpolation of a virtual pixel row is slight when a conversion coefficient including a low frequency component is generated at the edge of the image, but when a conversion coefficient including a high frequency component is generated. May cause significant image quality degradation at the end of the decoded image.

  In addition, as described with reference to FIG. 14, when JPEG2000 irreversible compression is performed, coefficients including the same frequency component in the same direction as the boundary are arranged at the boundary of the tile, and in particular, a high-frequency component is located at one of the tile boundaries. In some cases, the coefficients including are arranged in a line.

  In this case, pixels with significant image quality degradation due to pixel interpolation are arranged on the Tile division boundary surface, and as a result, a discontinuous surface is likely to be visually recognized on the Tile division boundary surface.

  The present invention has been made in view of the above problems, and it is a main object of the present invention to reduce visibility of image quality degradation by arranging coefficients including different frequency components on the Tile boundary surface in wavelet transform. To do.

The above objects, the two-dimensional image data is divided into a pixel area having a predetermined number of pixels, an image processing apparatus for performing a wavelet transform for each pixel area, the vertical direction of the pixel line or the lateral direction in the pixel area has a one-dimensional wavelet transform processing means in the pixel line unit performs one-dimensional wavelet transform, the one-dimensional wavelet transform process unit, the pixel lines as the high frequency components and low frequency components appear alternately 1 as well as the dimension wavelet transform, in adjacent pixels line, reverses the order of appearance of the high frequency and low frequency components have rows the conversion, for the same direction of the frequency component and the direction of the division boundary of the pixel area, the division boundary This is achieved by an image processing apparatus characterized in that coefficients of different frequency components are alternately arranged on the surface.

The above-described object is an image processing device that performs wavelet transform independently on two field images constituting one interlaced image, and is a vertical pixel line or horizontal pixel constituting field image data. 1-dimensional wavelet transform processing means for performing 1-dimensional wavelet transform in units of lines, and offset addition means for adding an odd-numbered offset to the horizontal address of each pixel for one of the field image data. For pixels that are even numbers, such that one of the high-frequency components or the low-frequency components appears for pixels whose horizontal addresses of the pixels included in the pixel line are odd. It converts the pixel line as the other high frequency components or low frequency components appear, from the two field images One of the different frequency components in a one-dimensional frequency component in the vertical direction in the interlaced image when constituting an interlaced image is also achieved by an image processing apparatus, characterized in that the alternately arranged.

The object described above is a two-dimensional image data is divided into a pixel area having a predetermined number of pixels, an image processing method for performing a wavelet transform for each pixel area, the vertical direction of the pixel lines in the pixel area or It has a one-dimensional wavelet transform processing step that performs one-dimensional wavelet transform in units of horizontal pixel lines, and the one-dimensional wavelet transform processing step has pixel lines so that high-frequency components and low-frequency components appear alternately. the addition to one-dimensional wavelet transform in the adjacent pixel line is reversed order of appearance of the high frequency and low frequency components have rows the conversion, the frequency components in the same direction as the direction of the division boundary of the pixel area, divided by the image processing method characterized in that coefficients of different frequency components in the plane of the boundary are alternately arranged It is made.

Further, the above-described object is an image processing method for performing wavelet transform independently on two field images constituting one interlaced image, in which vertical pixel lines or horizontal pixels constituting field image data are provided. A one-dimensional wavelet transform processing step for performing one-dimensional wavelet transform in units of lines, and an offset addition step for adding an odd offset to the horizontal address of each pixel with respect to one of the field image data. For pixels that are even, the transformation step is such that one of the high frequency components or low frequency components appears for pixels whose horizontal addresses of pixels included in the pixel line are odd. It converts the pixel line as the other high frequency components or low frequency components appear, the two fees Different frequency components in a one-dimensional frequency component in the vertical direction is also achieved by an image processing method characterized by alternately arranged at the time of constructing a single interlaced image from de image in interlaced image.

  The above-described object can also be achieved by a program that causes a computer to implement the method of the present invention, or a computer-readable recording medium that stores the program.

  According to the present invention, when the image is divided into a plurality of tiles and subjected to wavelet transform, the discontinuous surface at the end of the tile due to the tile division is performed by preventing the coefficients including the same frequency component from being arranged in a line. Visibility can be reduced.

Hereinafter, the present invention will be described in detail based on preferred embodiments with reference to the accompanying drawings.
● <First embodiment>
First, an image processing apparatus according to the first embodiment of the present invention will be described.
The image processing apparatus according to the first embodiment of the present invention inputs image data, and to the image data, a vertical pixel line from the upper end to the lower end and a horizontal line from the left end to the right end up to a predetermined decomposition level. A two-dimensional discrete wavelet transform is realized by repeatedly performing a one-dimensional lifting operation for each pixel line, and the calculated subband coefficients are output.

  The rough configuration of the apparatus will be described with reference to FIG. 1. The image processing apparatus according to the present embodiment switches the output destination of the operation switch 103 according to the coordinates of the one-dimensional pixel line, and is output from the buffer memory 102. A one-dimensional pixel column is processed by a first system called a pixel interpolator A104 to a one-dimensional lifting arithmetic unit A106, or a second system called a pixel interpolator B105 to a one-dimensional lifting arithmetic unit B107. A one-dimensional wavelet transform is performed to generate subband coefficients. Then, two-dimensional discrete wavelet transform is repeatedly performed on the coefficients of the subband (LL subband) including the LL component including the low frequency component in both the horizontal direction and the vertical direction until a predetermined decomposition level is reached.

  Here, the first system of the pixel interpolator A104 to the one-dimensional lifting arithmetic unit A106 and the first unit of the pixel interpolator B105 to the one-dimensional lifting arithmetic unit B107 performed by the image processing apparatus according to the first embodiment of the present invention. The processing contents of the second system will be described.

  When the image processing apparatus according to the present embodiment performs a lifting operation on an image (Tile image) divided into tiles, the order of the lifting operation depends on whether the vertical or horizontal pixel coordinates are even or odd. To change. In other words, when performing calculations for each vertical pixel line from the upper end to the lower end, according to the horizontal coordinates (line coordinates) of the pixel and for each horizontal pixel line from the left end to the right end Changes the order of lifting operations according to the vertical coordinates of the pixel.

  For this reason, two types of processing are performed: lifting computation when the pixel line coordinates are even, and pixel interpolation processing according to the computation, and the same processing when the pixel line coordinates are odd. That is, the image processing apparatus switches the output destination of the calculation switch 103 for each pixel line coordinate, and when the pixel line coordinate is an even number, the first line (pixel interpolators A104 to A104 through the calculation switch 103). When the one-dimensional lifting arithmetic unit A106) is used and the coordinates of the pixel line are odd numbers, the second system (pixel interpolator B105 to one-dimensional lifting arithmetic unit B107) is used in the same manner.

  First, processing in the first system (pixel interpolator A104 to one-dimensional lifting arithmetic unit A106) will be described. As described with reference to FIG. 11, the pixel interpolator A104 sets a predetermined number of pixels at the end of the pixel column depending on whether the coordinates of the start point and the end point of the pixel column are odd or even. Interpolate. In this embodiment, the number of pixels interpolated by the pixel interpolator A104 is 4 pixels when the coordinates are even, and 3 pixels when the coordinates are odd.

  A one-dimensional lifting operation is performed on the input pixel row in which the pixels are interpolated by a one-dimensional lifting operation unit A106. The recurrence formula of the lifting conversion here is as follows.

However,
α = −1.586 134 342 059 924
β = −0.052 980 118 572 961
γ = 0.882 911 075 530 934
δ = 0.443 506 852 043 971
K = 1.230 174 104 914 001
n is an integer
... (Formula 2)

  Next, processing in the second system (pixel interpolator B105 to 1-dimensional lifting arithmetic unit B107) will be described. The pixel interpolator B105 interpolates a predetermined number of pixels at the end of the pixel column depending on whether the coordinates of the start point and end point of the pixel column are odd or even. In the pixel interpolator B105, the number of pixels to be interpolated is 3 pixels when the coordinates of the pixels located at the ends are even, and 4 pixels when the coordinates are odd.

  The one-dimensional lifting arithmetic unit B107 performs one-dimensional lifting conversion on the input pixel string in which the pixels are interpolated. The recurrence formula of the lifting conversion here is as follows.

  The recurrence formula shown in (Formula 3) is the same as the recurrence formula shown in (Formula 2), but replaces the arithmetic processing when the pixel coordinates in the one-dimensional input pixel column are odd and the arithmetic processing when the pixel coordinates are even. Is. Therefore, as a result of the lifting calculation of (Equation 3), a high frequency component appears when the pixel coordinates are even (Y (2n)), and a low frequency component appears when the pixel coordinates are odd (Y (2n + 1)). That is, the operations of steps 2, 4, and 6 in (Equation 3) can be viewed as high-pass filter processing, and the operations of steps 1, 3, and 5 can be viewed as low-pass filter processing. It can be seen that the filtering process has been replaced.

  The image input to the image processing apparatus according to the first embodiment of the present invention is first subjected to a one-dimensional lifting operation from the left end to the right end for each vertical pixel line from the upper end to the lower end. In the lifting calculation for each vertical pixel line, (Equation 2) is used for pixel lines with an even horizontal coordinate from the left end to the right end, and (Equation 3) is used for odd pixel lines. The calculation result is shown in FIG. After performing the lifting operation for each vertical pixel line, in the wavelet transform, the coefficients are rearranged in order to compile the coefficients for the L and H components. The result of rearrangement is shown in FIG.

  The image processing apparatus according to the first embodiment of the present invention performs a one-dimensional lifting calculation from the upper end to the lower end for each horizontal pixel line following the lifting calculation for each vertical pixel line. Also for the lifting calculation for each horizontal pixel line, (Equation 2) is used for pixel lines with an even vertical coordinate from the upper end to the lower end, and (Equation 3) is used for odd pixel lines. The calculation result is shown in FIG. After performing the lifting operation for each horizontal pixel line, in the wavelet transform, the coefficients are rearranged in order to collect the coefficients for the L and H components. The result is shown in FIG.

  FIG. 3 shows where the subband coefficients generated by applying the wavelet transform according to the first embodiment to an image of 4 pixels vertically and 4 pixels horizontally are located in the original image. FIG. 3A shows the distribution of coefficients generated by one-dimensional wavelet transform for each vertical pixel line from the upper end to the lower end in the original positional relationship, and FIG. 3B shows the first embodiment. 2 shows the distribution of coefficients generated by the two-dimensional wavelet transform according to the positional relationship of the original pixels.

  As is clear from the comparison with FIG. 13, the processing result by the image processing apparatus of this embodiment shows that the frequency component of the coefficient generated by the wavelet transform is adjacent to the one-dimensional frequency component in the vertical direction or the horizontal direction. It turns out that it is a component different from the frequency component of a coefficient.

Next, details of the configuration of the image processing apparatus according to the present embodiment will be described with reference to FIG.
The buffer memory 102 inputs image data to be wavelet transformed based on an image data input command from the CPU 101. The buffer memory 102 outputs the data string to the arithmetic switch 103 based on the data string output command from the CPU 101. Further, the subband coefficients output from the one-dimensional lifting calculator A106 or the one-dimensional lifting calculator B107 are temporarily stored. In addition, the buffer memory 102 outputs all the subband coefficients that have been calculated based on the coefficient output command from the CPU 101.

The arithmetic switching unit 103 selectively outputs the input one-dimensional data string to the pixel interpolator A104 or the pixel interpolator B105 based on the switching information from the CPU 101.
The pixel interpolator A104 inputs a one-dimensional data string and the coordinates of the start point and the end point of the data, and determines whether the virtual data string necessary for the one-dimensional lifting operation based on (Equation 2) is an odd number of coordinates of the start point and the end point. Depending on whether it is an even number, it is interpolated at both ends of the data string and output to the one-dimensional lifting calculator A106. The pixel interpolator B105 interpolates the virtual data string necessary for the one-dimensional lifting calculation based on (Equation 3) at both ends of the data string in accordance with the coordinates of the start point and the end point, and sends it to the one-dimensional lifting calculator B107. Output.

  The one-dimensional lifting arithmetic unit A106 inputs a one-dimensional data string, performs a one-dimensional lifting operation based on the recurrence formula shown in (Equation 2), and generates a low frequency for the generated one-dimensional subband coefficient. The coefficients including the components are rearranged so that the coefficients including the high-frequency components are concentrated on the left side and the coefficients including the high-frequency component are aligned on the right side, and are output to the buffer memory 102. The one-dimensional lifting calculator B107 performs the same processing as described above based on the recurrence formula shown in (Expression 3).

  First, the CPU 101 first sets the image size information of the input image data and the information of the decomposition level number of the wavelet transform, for example, by setting from a non-illustrated non-volatile memory, an unillustrated input device or the like, or an unillustrated external circuit or external After being acquired from the apparatus, in order to perform two-dimensional wavelet transform on the image data, the above-described various information and commands are output to the buffer memory 102, the operation switching unit 103, the pixel interpolator A104, and the pixel interpolator B105.

Next, the operation of the image processing apparatus having the above configuration will be described.
FIG. 4 is a flowchart for explaining the operation of the image processing apparatus according to the first embodiment of the present invention.
First, in step S401, the CPU 101 acquires image size information of input image data and information on the number of wavelet transform decomposition levels. Based on the image size information, the CPU 101 transmits a memory allocation command and an image data input command for the image size to the buffer memory 102. The buffer memory 102 secures an area for the image size based on the memory securing command transmitted from the CPU 101, and then receives the image data input command and acquires the image data from an external storage device such as a hard disk. Temporarily store in the reserved area.

  Next, in step S402, the CPU 101 sequentially reads the coordinates from the left end to the right end (lateral direction) of the image. Each time the coordinates are read, the CPU 101 instructs the calculation switch 103 of the output destination of the input data.

Here, an operation when the coordinates read by the CPU 101 are an even number and an operation when the coordinates are an odd number will be described.
If the read horizontal pixel coordinate is an even number, the CPU 101 instructs the arithmetic switching unit 103 to output the pixel row from the buffer memory 102 to the first system (pixel interpolator A104). The arithmetic switch 103 switches a built-in switch so that input data is output to the pixel interpolator A104 based on a command from the CPU 101. Thereafter, the CPU 101 transmits the vertical coordinate at the start point and the end point of the pixel column to the pixel interpolator A104 for the vertical pixel line from the upper end to the lower end corresponding to the read horizontal pixel coordinate. Thereafter, the CPU 101 instructs the buffer memory 102 to output the vertical pixel row. The buffer memory 102 outputs the designated pixel row to the arithmetic switch 103 based on the command transmitted from the CPU 101.

  The arithmetic switching unit 103 outputs the pixel sequence transmitted from the buffer memory 102 to the pixel interpolator A104 according to the setting. When the pixel column is input, the pixel interpolator A104, based on the coordinates of the start point and end point of the pixel column transmitted from the CPU 101, for example, the coordinates of the pixel located at the end are even numbers as described with reference to FIG. In this case, 4 pixels are interpolated at both ends of the pixel column and output in the one-dimensional lifting arithmetic unit A106. When a pixel column is input, the one-dimensional lifting calculator A106 performs a one-dimensional lifting operation on the entire pixel column based on the recurrence formula of (Equation 2), and the subband coefficient generated by the calculation is calculated for each subband. And output to the buffer memory 102. The buffer memory 102 temporarily stores the input subband coefficients.

  If the read horizontal pixel coordinate is an odd number, the CPU 101 instructs the arithmetic switching unit 103 to output the pixel row from the buffer memory 102 to the second system (pixel interpolator B105). The arithmetic switch 103 switches a built-in switch so that input data is output to the pixel interpolator B105 based on a command from the CPU 101. Thereafter, the CPU 101 transmits the vertical coordinate at the start point and the end point of the pixel column to the pixel interpolator B105 for the vertical pixel line from the upper end to the lower end corresponding to the read horizontal pixel coordinate. Thereafter, the CPU 101 instructs the buffer memory 102 to output the vertical pixel row. The buffer memory 102 outputs the designated pixel row to the arithmetic switch 103 based on the command transmitted from the CPU 101.

  The arithmetic switch 103 outputs the pixel string transmitted from the buffer memory 102 to the pixel interpolator B105 according to the setting. When the pixel column is input, the pixel interpolator B105, based on the coordinates of the start point and end point of the pixel column transmitted from the CPU 101, is 3 pixels when the coordinates of the pixels located at the end are even, and 4 when the coordinates are odd. Only the pixels are interpolated and output to the one-dimensional lifting calculator B107. When the pixel column is input, the one-dimensional lifting calculator B107 performs a one-dimensional lifting operation on the entire pixel column based on the recurrence formula of (Equation 3), and the subband coefficients generated by the calculation are subbands. The data are rearranged every time and output to the buffer memory 102. The buffer memory 102 temporarily stores the input subband coefficients.

In step S402, the series of processes as described above are repeated until the CPU 101 finishes counting all the pixel coordinates in the horizontal direction from the left end to the right end of the image.
Here, the subband coefficients temporarily stored in the buffer memory 102 when the processing of the horizontal one-dimensional wavelet transform performed from the left end to the right end for each vertical pixel column is completed are the one-dimensional lifting calculator A106. Alternatively, since the rearrangement is performed for each subband in the one-dimensional lifting calculator B107, the coefficients are temporarily stored in a state where the coefficients are collected for each subband as shown in FIG.

  Next, in step S403, the processing in step S402 performed for the vertical pixel line from the upper end to the lower end for each coordinate of the horizontal pixel line is performed from the left end to the right end for each coordinate of the vertical pixel line from the upper end to the lower end. The same is done for the pixel columns in the direction.

  Here, the subband coefficients temporarily stored in the buffer memory 102 when the vertical one-dimensional wavelet transform performed from the upper end to the lower end for each coordinate of the horizontal pixel line is completed are the subband coefficients for each subband. Since the data are rearranged and stored, the coefficients are temporarily stored in a state where the coefficients are collected for each subband as shown in FIG.

  Next, in step S404, the CPU 101 determines whether or not a predetermined decomposition level has been reached by two-dimensional wavelet transformation based on the information on the number of decomposition levels acquired in step S401. If it is determined that the predetermined decomposition level has been reached, the process proceeds to step S405, and all the subband coefficients temporarily stored in the memory buffer 102 are output. If it is not determined that the predetermined decomposition level has been reached, among the subband coefficients temporarily stored in the memory buffer 102, the LL subband including the low frequency component in both the horizontal direction and the vertical direction is again determined. Steps S402 and S403 are performed.

  Here, the coefficient distribution when the above-described wavelet transform is applied to Tile will be described. FIG. 5 shows the distribution of wavelet transform coefficients generated by the image processing apparatus according to the present embodiment when the image of 4 pixels vertically and 8 pixels horizontally is divided into two tiles of 4 pixels vertically and horizontally. This is indicated by the positional relationship. FIG. 5A shows the distribution of coefficients generated by the two-dimensional wavelet transform by the image processing apparatus according to the present embodiment. FIG. 5 (b) shows only the distribution of frequency components in the same vertical direction as the direction of the tile division boundary among the coefficients shown in FIG. 5 (a). From FIG. 5B, it can be seen that when Tile division is performed, coefficients of different frequency components are alternately arranged on the division boundary surface with respect to frequency components in the same vertical direction as the direction of the Tile division boundary.

  By performing the above operation, when the image processing apparatus according to the first embodiment of the present invention performs JPEG2000 encoding on an image divided into a plurality of tiles, the coordinates of the pixel lines in the vertical and horizontal directions are changed. By changing the order of the lifting operation depending on whether it is an even number or an odd number, it is possible to prevent the coefficients including the same frequency component from being arranged in a line.

● <Second Embodiment>
The image processing apparatus according to the second embodiment of the present invention will be described below.
The image processing apparatus according to the present embodiment inputs a first field image or a second field image constituting an interlaced image, and compresses one field image with respect to one field image when the field image is independently compressed by the JPEG2000 method. By applying a JPEG2000 format compression by applying a horizontal offset to a coordinate system called a reference grid defined in the JPEG2000 format, when an interlaced image is constructed from a field image, the interlaced image moves from the upper end to the lower end. With respect to the one-dimensional frequency component in the vertical direction, coefficients including the same frequency component are prevented from being arranged in a line.

  Here, a rough configuration of the image processing apparatus according to the present embodiment will be described with reference to FIG. In FIG. 6, the same components as those in FIG. 1 are denoted by the same reference numerals. In the image processing apparatus according to the present embodiment, the one-dimensional pixel row output from the buffer memory 102 is converted by the pixel interpolator A104 to the one-dimensional lifting arithmetic unit A106 based on the coordinate information of the input image output from the CPU 601. By being processed, one-dimensional wavelet transform is performed on the pixel row, and subband coefficients are generated. When the field image to which the CPU 601 gives an offset to the reference grid is an input image, the coordinate information to which the offset is given is transmitted from the CPU 601 to the pixel interpolator A104 and the one-dimensional lifting calculator A106. The image processing apparatus according to the present embodiment repeatedly performs two-dimensional wavelet transform on subband coefficients including LL components until a predetermined decomposition level is reached.

  This process will be described in detail using an interlaced image of 8 pixels vertically and 4 pixels horizontally shown in FIG. FIG. 7 is a schematic diagram for explaining a process of assigning only one lateral offset to the reference grid of the second field image, performing two-dimensional wavelet transform, and dividing the image into subbands. FIG. 7A is an interlaced image of an original image composed of input field images. Moreover, the 1st field image and 2nd field image which were divided | segmented from the interlace image are shown in FIG.7 (b). Note that “original image line No.” in FIG. 7 is a vertical address number from the upper end to the lower end of the interlaced image. The numbers shown in parentheses represent the coordinates of the pixel after offset addition in each field image, the left value is the vertical coordinate from the upper end to the lower end, and the right value is from the left end to the right end. Represents horizontal coordinates. As shown in FIG. 7B, the second field image is given an offset of +1 in the horizontal direction.

  The upper part of FIG. 7B is a first field image constituting the interlaced image (frame image) of FIG. 7A, and the lower part is a second field image. The predetermined frame image input to the image processing apparatus is subjected to a one-dimensional lifting operation using (Equation 2) from the left end to the right end for each vertical pixel line from the upper end to the lower end for each field image. . The calculation result is shown in FIG. After performing the lifting operation for each vertical pixel line, in the wavelet transform, the coefficients are rearranged in order to compile the coefficients for the L and H components. The result of the rearrangement is shown in FIG.

  The image processing apparatus according to the present embodiment performs a one-dimensional lifting calculation using (Equation 2) from the upper end to the lower end for each horizontal pixel line following the lifting calculation for each vertical pixel line. The calculation result is shown in FIG. Here, for the first field image, since no offset is given to the coordinate value of the reference grid, the horizontal address from the left end to the right end starts with 0, that is, an even number. Therefore, regarding the subband coefficients of the first field image generated by the lifting operation, the vertical one-dimensional frequency component included in the coefficients is L, H, L,. Lined up. On the other hand, for the second field image, since the offset value 1 is given in the horizontal direction, the horizontal address starts with 1, that is, an odd number. Therefore, with respect to the subband coefficients of the second field image generated by the lifting operation, the vertical one-dimensional frequency components included in the coefficients are H, L, H,. line up.

  Therefore, it can be seen from FIG. 7 (e) that even when the images have the same size, the subband coefficients generated by the lifting operation differ depending on whether the offset is applied to the reference grid. After performing the lifting operation for each horizontal pixel line, in the wavelet transform, the coefficients are rearranged in order to collect the coefficients for the L and H components. The result is shown in FIG.

  FIG. 8 shows where the conversion coefficients of the first and second fields shown in FIG. 7 are located in the original frame image of 8 pixels vertically and 4 pixels horizontally. FIG. 8A shows the distribution of coefficients generated by one-dimensional wavelet transform (in the horizontal direction) for each vertical pixel line from the upper end to the lower end in the original positional relationship. Similarly, FIG. 8B shows the distribution of coefficients generated by the two-dimensional wavelet transform in the positional relationship of the original pixels.

  As shown in FIG. 7E, in the first field image to which no offset is given, the subband coefficients are L, H, L, H and L starting from the left end to the right end for each vertical pixel line. In the second field image to which the array and offset are added, since the subband coefficients are arranged with H, L, H, L, and H as the head from the left end to the right end for each vertical pixel line, the first field image L and H appear alternately in the one-dimensional frequency component in the vertical direction in the interlaced image in FIG. 8B in which the second field image is arranged on the odd lines. From these facts, as shown in FIG. 8, regarding the one-dimensional frequency component in the vertical direction from the upper end to the lower end, the frequency component of the coefficient generated by the wavelet transform is different from the frequency component of the adjacent coefficient. I understand.

  In the above description, only one lateral offset is given to the reference grid. However, the present invention is not limited to this, and the offset may be any positive odd number. Further, an offset may be given to the first field image instead of the second field image. If the difference in offset is an odd number, an offset may be given to both field images.

  Next, details of the configuration of the image processing apparatus according to the second embodiment of the present invention will be described with reference to FIG. In FIG. 6, components having functions equivalent to those of the first embodiment of the present invention are denoted by the same numbers.

  The buffer memory 102 inputs image data to be wavelet transformed based on an image data input command from the CPU 601. The buffer memory 102 outputs the data string to the pixel interpolator A 104 based on the data string output command from the CPU 601. In addition, the subband coefficients output from the one-dimensional lifting calculator A106 are temporarily stored. Further, the buffer memory 102 outputs all the subband coefficients for which computation has been completed based on the coefficient output command from the CPU 601.

  The pixel interpolator A104 and the one-dimensional lifting calculator A106 perform the same processing as in the first embodiment, and perform one-dimensional wavelet transformation on the data string output from the buffer memory 102.

  First, the CPU 601 determines image size information of input image data, information on the number of wavelet transform decomposition levels, and field classification information for determining whether the input image is a first field image or a second field image. Enter. In order to perform two-dimensional wavelet transform on the image data, the CPU 601 outputs the above-described various information and commands to the buffer memory 102, the operation switching unit 103, the pixel interpolator A104, and the pixel interpolator B105, respectively. If the offset of the reference grid of the input image is a field image to which a value is given, the CPU 601 gives a predetermined offset value to the horizontal coordinate in the reference grid of the image.

  Next, the operation of the image processing apparatus according to the present embodiment will be described using the flowchart shown in FIG. In FIG. 9, steps that perform the same processing as in the first embodiment are denoted by the same numbers. In the following description, the offset to be given to the horizontal coordinate of the reference grid is 1 and the field image to which the offset is given is only the second field image in combination with the above explanation.

  First, in step S901, the CPU 601 determines the image size information of the image data, the information of the wavelet transform decomposition level number, and the field classification for determining whether the input image is the first field image or the second field image. Get information. Based on the image size information, the CPU 601 transmits a memory allocation command for the image size and an input command for image data to the buffer memory 102. The buffer memory 102 secures an area for the image size based on the memory securing command transmitted from the CPU 601, and then receives the image data input command and inputs the image data.

Next, in step S <b> 902, the CPU 601 determines whether the acquired field classification information indicates a first field image or a second field image. If the field division information indicates the second field division image, the process proceeds to step S903. If the input field division information indicates the first field division image, the process proceeds to step S904.
In step S903, the CPU 601 gives the offset value 1 to the horizontal coordinate of the reference grid of the second field image.

  In step S904, the CPU 601 sequentially reads horizontal pixel coordinates from the left end to the right end of the image size. Further, the CPU 601 uses the pixel interpolator A104 and the one-dimensional lifting calculator A106 as the vertical coordinates at the start point and the end point of the pixel row for the vertical pixel row from the upper end to the lower end corresponding to the read horizontal pixel coordinate. Send to. Thereafter, the CPU 601 transmits to the buffer memory 102 a command for outputting a vertical pixel row corresponding to the read horizontal pixel coordinates.

  The buffer memory 102 outputs the designated pixel row to the pixel interpolator A104 based on the command transmitted from the CPU 601. When a pixel column is input to the pixel interpolator A104, the pixel interpolator A104, based on the coordinate values of the start point and end point of the pixel column transmitted from the CPU 601, for example, at the end as described with reference to FIG. When the coordinates of the pixels located are even numbers, four pixels are interpolated at both ends of the pixel column, and when the coordinates are odd, three pixels are interpolated and output to the one-dimensional lifting calculator A106. When the pixel column is input, the one-dimensional lifting calculator A106 performs a one-dimensional lifting operation on the entire pixel column based on the recurrence formula of (Equation 2), and the subband coefficients generated by the calculation are further subbands. The data are rearranged every time and output to the buffer memory 102. The buffer memory 102 temporarily stores the input subband coefficients.

  In step S904, the series of processes as described above are repeated until the CPU 601 finishes counting all the horizontal coordinates from the left end to the right end of the image. The result is shown in FIG. In FIG. 7C, the upper diagram is the first field image, and the lower diagram is the second field image.

  Here, FIG. 7D shows the state of the subband coefficients temporarily stored in the buffer memory 102 when the horizontal one-dimensional wavelet transform performed from the left end to the right end is completed for each vertical pixel column. Shown in In FIG. 7D, the upper diagram is a first field image, and the lower diagram is a second field image. Since the subband coefficients are rearranged for each subband in the one-dimensional lifting calculator A106, the subband coefficients are temporarily stored in a state where the coefficients are collected for each subband.

  Next, in step S905, the processing in step S904 performed for the vertical pixel column from the upper end to the lower end for each horizontal pixel line coordinate is the same for the horizontal pixel column from the left end to the right end for each vertical pixel coordinate. To do. The result is shown in FIG. In FIG. 7E, the upper diagram is a first field image, and the lower diagram is a second field image.

  Here, FIG. 7 (f) shows the subband coefficients temporarily stored in the buffer memory 102 when the vertical one-dimensional wavelet transform performed from the upper end to the lower end for each horizontal pixel column is completed. . Since the subband coefficients are rearranged for each subband and stored, the subband coefficients are temporarily stored in a state where the coefficients are collected for each subband. In FIG. 7F, the upper diagram is the first field image, and the lower diagram is the second field image.

  Next, in step S404, as in the first embodiment, the CPU 601 determines whether or not a predetermined decomposition level has been reached by two-dimensional wavelet transformation based on the information on the number of decomposition levels acquired in step S901. to decide. If it is determined that the predetermined decomposition level has been reached, the process proceeds to step S405, and all the subband coefficients temporarily stored in the memory buffer 102 are output. If it is not determined that the predetermined decomposition level has been reached, among the subband coefficients temporarily stored in the memory buffer 102, the LL subband including the low frequency component in both the horizontal direction and the vertical direction is again determined. Steps S904 and S905 are performed.

  Here, the distribution of coefficients when the wavelet transform performed by the image processing apparatus according to the present embodiment is applied to Tile will be described using an interlaced image of 8 pixels in the vertical direction and 8 pixels in the horizontal direction. FIG. 10 shows the distribution of coefficients generated by the wavelet transform performed in this embodiment in the positional relationship of the original pixels when this interlaced image is divided into two tiles of 8 pixels in the vertical direction and 4 pixels in the horizontal direction. It is a thing. FIG. 10A shows the distribution of coefficients generated by the two-dimensional wavelet transform in this embodiment. FIG. 10B shows only the distribution of frequency components in the same vertical direction as the direction of the Tile division boundary among the coefficients shown in FIG. From FIG. 10B, it can be seen that when Tile division is performed, coefficients of different frequency components are alternately arranged on the division boundary surface with respect to frequency components in the same vertical direction as the direction of the Tile division boundary.

  As described above, when the image processing apparatus according to the present embodiment inputs the first field image or the second field image constituting the interlaced image and independently encodes each of them, A one-dimensional frequency component in the vertical direction from the upper end to the lower end by adding an offset to at least one field image so that the difference in offset in the horizontal direction is a positive odd number in the Reference grid and performing JPEG2000 encoding. , It is possible to prevent the coefficients including the same frequency component from being arranged in a line.

<Other embodiments>
In the above-described embodiment, only the image processing apparatus including one device has been described. However, an equivalent function may be realized by a system including a plurality of devices.
A software program that realizes the functions of the above-described embodiments is supplied directly from a recording medium or to a system or apparatus having a computer that can execute the program using wired / wireless communication. The present invention includes a case where an equivalent function is achieved by a computer executing the supplied program.

Accordingly, the program code itself supplied and installed in the computer in order to implement the functional processing of the present invention by the computer also realizes the present invention. That is, the computer program itself for realizing the functional processing of the present invention is also included in the present invention.
In this case, the program may be in any form as long as it has a program function, such as an object code, a program executed by an interpreter, or script data supplied to the OS.

  As a recording medium for supplying the program, for example, a magnetic recording medium such as a flexible disk, a hard disk, a magnetic tape, MO, CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-R, DVD- There are optical / magneto-optical storage media such as RW, and non-volatile semiconductor memory.

As a program supply method using wired / wireless communication, a computer program forming the present invention on a server on a computer network, or a computer forming the present invention on a client computer such as a compressed file including an automatic installation function A method of storing a data file (program data file) that can be a program and downloading the program data file to a connected client computer can be used. In this case, the program data file can be divided into a plurality of segment files, and the segment files can be arranged on different servers.
That is, the present invention includes a server device that allows a plurality of users to download a program data file for realizing the functional processing of the present invention on a computer.

  In addition, the program of the present invention is encrypted, stored in a storage medium such as a CD-ROM, distributed to the user, and key information for decrypting the encryption for a user who satisfies a predetermined condition is provided via a homepage via the Internet, for example. It is also possible to realize the program by downloading it from the computer and executing the encrypted program using the key information and installing it on the computer.

  In addition to the functions of the above-described embodiments being realized by the computer executing the read program, the OS running on the computer based on the instruction of the program is a part of the actual processing. Alternatively, the functions of the above-described embodiment can be realized by performing all of them and performing the processing.

  Furthermore, after the program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion board or The CPU of the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments can also be realized by the processing.

1 is a block diagram illustrating a configuration example of an image processing apparatus according to a first embodiment of the present invention. It is a figure explaining the subband division | segmentation by the image processing apparatus which concerns on the 1st Embodiment of this invention. FIG. 3 is a schematic diagram showing where each subband coefficient is located in the original image in the subband division of FIG. 2. 3 is a flowchart showing an operation of the image processing apparatus according to the first embodiment of the present invention. FIG. 6 is a schematic diagram illustrating where in a source image a frequency component included in a wavelet transform coefficient is located when an image that has been tile-divided in the vertical direction is processed by the image processing apparatus according to the first embodiment. It is a block diagram which shows the structural example of the image processing apparatus which concerns on the 2nd Embodiment of this invention. It is a figure explaining the subband division | segmentation by the image processing apparatus which concerns on the 2nd Embodiment of this invention. FIG. 8 is a schematic diagram illustrating where each coefficient of a subband is located in the original image in the subband division of FIG. 7. It is a flowchart which shows operation | movement of the image processing apparatus which concerns on the 2nd Embodiment of this invention. FIG. 10 is a schematic diagram illustrating where in a source image a frequency component included in a wavelet transform coefficient is located when an image that has been tile-divided in the vertical direction is processed by the image processing apparatus according to the second embodiment. It is a schematic diagram explaining the pixel interpolation method of a JPEG2000 system. It is a figure explaining the subband division | segmentation in the wavelet transformation of a JPEG2000 system. It is a schematic diagram which shows where each coefficient of a subband is located in the original image in the subband division | segmentation of FIG. It is a schematic diagram which shows where the frequency component contained in a wavelet transform coefficient is located in the original image when the image divided into the tile in the vertical direction is processed based on the JPEG2000 system.

Claims (8)

An image processing apparatus that divides two-dimensional image data into pixel areas having a predetermined number of pixels and performs wavelet transform for each pixel area ,
Has a one-dimensional wavelet transform processing means for performing one-dimensional wavelet transform on a pixel line basis in the vertical direction of the pixel line or the lateral direction that constitute the pixel area,
The one-dimensional wavelet transform processing means is
The pixel line is subjected to a one-dimensional wavelet transform so that a high frequency component and a low frequency component appear alternately,
In the adjacent pixel lines, wherein the high-frequency component reverses the order of appearance of the low-frequency components have rows the conversion, for the same direction of the frequency component and the direction of the division boundary of the pixel area, the surface of the division boundary An image processing apparatus , wherein coefficients of different frequency components are alternately arranged .   The calculation performed by the one-dimensional wavelet transform processing means is composed of a first calculation for obtaining a high-frequency component and a second calculation for obtaining a low-frequency component, and the one-dimensional wavelet transform processing means The image processing apparatus according to claim 1, wherein an operation in which the order of the first operation and the second operation is switched is performed for each pixel line. An image processing apparatus that performs wavelet transform independently on two field images constituting one interlaced image,
One-dimensional wavelet transform processing means for performing one-dimensional wavelet transform in units of vertical pixel lines or horizontal pixel lines constituting the field image data;
Offset adding means for adding an odd offset to the horizontal address of each pixel with respect to one of the field image data;
The one-dimensional wavelet transform processing means is
The high frequency component is applied to a pixel that is an even number, such that one of a high frequency component or a low frequency component appears for a pixel that has an odd horizontal address in the pixel line. Alternatively, when the pixel line is transformed so that the other of the low frequency components appears and one interlaced image is constructed from the two field images, the frequency differs in the one-dimensional frequency component in the vertical direction in the interlaced image. An image processing apparatus, wherein components are arranged alternately . An image processing method for dividing two-dimensional image data into pixel areas having a predetermined number of pixels and performing wavelet transform for each pixel area ,
Has a one-dimensional wavelet transform processing step of performing one-dimensional wavelet transform on a pixel line basis in the vertical direction of the pixel line or the lateral direction that constitute the pixel area,
The one-dimensional wavelet transform processing step includes:
The pixel line is subjected to a one-dimensional wavelet transform so that a high frequency component and a low frequency component appear alternately,
In the adjacent pixel lines, wherein the high-frequency component reverses the order of appearance of the low-frequency components have rows the conversion, for the same direction of the frequency component and the direction of the division boundary of the pixel area, the surface of the division boundary An image processing method, wherein coefficients of different frequency components are alternately arranged .   The calculation performed by the one-dimensional wavelet transform processing step is composed of a first calculation for obtaining a high-frequency component and a second calculation for obtaining a low-frequency component, and the one-dimensional wavelet transform processing step comprises The image processing method according to claim 4, wherein an operation in which an order of the first operation and the second operation is exchanged is performed for each pixel line. An image processing method for performing wavelet transform independently on two field images constituting one interlaced image,
A 1-dimensional wavelet transform processing step of performing one-dimensional wavelet transform on a pixel line basis in the vertical direction of the pixel line or lateral constituting the field image data,
An offset addition step of adding an odd offset to the horizontal address of each pixel with respect to one of the field image data;
The one-dimensional wavelet transform processing step includes:
The high frequency component is applied to a pixel that is an even number, such that one of a high frequency component or a low frequency component appears for a pixel that has an odd horizontal address in the pixel line. Alternatively, when the pixel line is transformed so that the other of the low frequency components appears and one interlaced image is constructed from the two field images, the frequency differs in the one-dimensional frequency component in the vertical direction in the interlaced image. An image processing method, wherein components are arranged alternately . The program for making a computer perform each process as described in the image processing method of any one of Claim 4 thru | or 6. A computer-readable recording medium on which the program according to claim 7 is recorded.

Images