JPH0846966A - Method for improving picture quality - Google Patents

Abstract

PURPOSE:To reduce mosquito noise in the neighborhood of the contour of an encoding picture by a DCT encoding system. CONSTITUTION:A deteriorated picture caused by encoding is filter-processed by a second differentiation circuit 35 and a picture element value taking the absolute value of the value of the processing is generated. The signal is given to a degree distribution calculator 37 and the picture element degree distribution of a second differential picture. Moreover, a threshold value for deviding the picture element degree distribution into two distributions is calculated by a threshold value calculator 39. Through the use of the threshold value, a comparator 40 and a block converting equipment 42 separates a block with large luminance change from the block with slight luminance change. A binary picture block which does not constitute the connection contour or the texture contour within the separated blocks is adopted as a distortion block. Then, the distortion block is smoothed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image quality improving method for an image, and more particularly, when a decoder reproduces an image compression-encoded by an encoding method using DCT (discrete cosine transform), a reproduced image is reproduced. The present invention relates to an image quality improving method for reducing the distortion generated near the contour of the image.

[0002]

2. Description of the Related Art In recent years, images have been digitally compressed and encoded,
It is often transmitted or recorded. Currently, one of the most popular image compression encoding methods is an encoding method using DCT. An image distortion peculiar to this method is a distortion called mosquito noise that occurs near the contour of an image. This is a hazy distortion that occurs near the contour (edge) of the image. Especially in the case of a moving image, it looks as if a mosquito is flying along the contour, which is very annoying. It is considered that this distortion occurs as a leak to the low frequency because of the coarse quantization in the middle and high frequencies of the DCT coefficient. If a two-dimensional area in which the brightness and hue of the image are not significantly changed is a flat area, mosquito noise in the flat area near the contour causes a large visual deterioration in image quality. Therefore, if this distortion can be reduced, the image quality is greatly improved.

Conventional methods for reducing mosquito noise are roughly classified into a method of reducing mosquito noise by an encoder and a method of reducing mosquito noise by a decoder. A typical method for reducing mosquito noise in an encoder is to detect an area where mosquito noise is likely to occur in advance and reduce the quantization width of the DCT coefficient in the image area to reduce the noise. Is the way. On the other hand, a typical method of reducing mosquito noise on the decoder side is to detect a contour portion and smooth the neighboring pixels to reduce the noise.

[0004]

However, such conventional methods have the following problems, respectively. First, the conventional method of reducing the mosquito noise in the encoder cannot be used if the transmission bit rate is limited to a low value in the encoder and the quantization width cannot be further reduced. Moreover, the noise is actually likely to occur when the transmission bit rate is considerably low. This limits the use of this method depending on the transmission bit rate.

On the other hand, the method of reducing mosquito noise in the decoder is not affected by the transmission bit rate, and the noise can be reduced only by the processing on the decoder side. However, the conventional method in the decoder includes a contour (hereinafter, referred to as a texture contour) that constitutes a fine texture (for example, a fine image in which the change in grayscale is not large like a fiber) whose luminance change is not so large. Such images are not sufficiently considered. Originally, it is difficult to correctly distinguish the original contour of an image from the false contour caused by noise. Especially, it is more difficult to discriminate the image including the texture contour.

If the noise-reduced portion is excessively widened in an attempt to remove noise by using a smoothing filter, the texture contour is also smoothed, resulting in a blurred and sharp image, which rather deteriorates the image quality. Moreover, the mosquito noise and the texture contour often have the same brightness change. For this reason, it has been considered to use an adaptive smoothing filter that changes the degree of smoothing depending on changes in the brightness of an image, a smoothing filter, a median filter, an emphasis filter, or the like for each image region. However, even with these conventional correction methods, if the noise and the texture are not properly discriminated, side effects such as blurring or distorting the texture contour or enhancing the noise are caused. Therefore, the conventional method has a drawback in that even though the noise is reduced, the texture portion is also blurred or distorted at the same time, and a sufficient effect cannot be obtained in terms of overall image quality.

[0007] Here, it is difficult to distinguish whether the mosquito noise generated near the texture contour in the image is the original contour or the pseudo contour due to the noise, and it is often visually inconspicuous. Further, when mosquito noise generated near the texture contour is filtered by a smoothing filter or a median filter, the texture contour is rather blurred or distorted, and is often visually noticeable.

The present invention has been made in view of such conventional problems, and a method of reducing mosquito noise by a decoder is used to protect texture contours included in an image, and only the noise is protected. Reduce. That is, the image quality is improved by avoiding the filtering process in the texture contour portion and smoothing only the noise that is generated in the portion where the mosquito noise is most noticeable visually, that is, the flat portion near the strong contour where the luminance change is large. The object is to realize an image improvement method capable of achieving

[0009]

According to a first aspect of the present invention, in the vicinity of a contour of a deteriorated image in which an original image composed of a plurality of pixel data is deteriorated, image distortion occurs in a flat portion of luminance near the contour. An image quality improving method for improving the deterioration of a deteriorated image into blocks having a predetermined number of pixels to obtain a plurality of deteriorated image blocks, and for each deteriorated image block obtained in the block dividing step. , A variation value calculation step for calculating the variation of pixel data in the block, and a frequency distribution calculation for calculating the frequency distribution of the variation values of the deteriorated image block for the unit screen using the variation value obtained in the variation value calculation step as a class value. 2 for dividing the frequency distribution obtained in the step and the frequency distribution calculation step into three according to the strength of the image contour.
A threshold value calculation step of calculating a first threshold value that is one class value and a second threshold value that is a threshold value smaller than the first threshold value;
A first extraction step of extracting, as a first block, a deteriorated image block whose lower limit is the first threshold obtained in the threshold calculation step, and a deteriorated image whose upper limit is the second threshold obtained in the threshold calculation step. A second extraction step of extracting the block as a second block; an expansion block calculation step of expanding the first block obtained in the first extraction step to a predetermined size and calculating an expansion block; Of the second blocks obtained in the second extraction step, the distortion block determination step of determining a block whose area overlaps with the expansion block obtained in the expansion block calculation step as a distortion block and the distortion block determination step A smoothing step of smoothing each pixel in the distortion block to obtain an image with reduced noise. A.

The invention of claim 3 of the present application is an image quality improving method for improving the distortion generated in the flat portion of the brightness near the contour in the vicinity of the contour of the degraded image in which the original image composed of a plurality of pixel data is degraded. Therefore, the deteriorated image is filtered by a secondary differential filter to be an absolute value, and a secondary differential image calculation step of calculating a secondary differential image and a secondary differential image obtained in the secondary differential image calculation step are performed. 2 for each pixel
Using the secondary differential value as a class value, the frequency distribution calculation step of calculating the frequency distribution of the secondary differential value for the unit screen, and the frequency distribution obtained in the frequency distribution calculation step are divided into two distributions according to the strength of the image contour. Therefore, a threshold value calculation step of calculating a specific class value as a threshold value, and a pixel of a secondary differential image having a threshold value obtained as a lower limit in the threshold value calculation step as a first value, and a secondary differential image having a threshold value as an upper limit Is used as a second value, the unit screen is replaced with a binarized image, and the binarized image obtained in the binarized step is divided into blocks each having a predetermined number of pixels. Of the binarized image blocks obtained by the block dividing step and the binarized image blocks obtained in the block dividing step.
And a smoothing block for discriminating, as a distorted block, a binarized image block in which the number of pixels of the first value is equal to or less than a predetermined number among the blocks obtained in the counting step. A determination step,
And a smoothing step of smoothing each pixel in the distortion block obtained in the smoothing block determination step to obtain an image with reduced noise.

According to a fourth aspect of the present invention, an image quality improving method for improving an image distortion generated in a flat portion of luminance near the contour in the vicinity of the contour of the deteriorated image in which the original image composed of a plurality of pixel data is degraded. In addition, when the deteriorated image includes a moving image composed of a plurality of frames, a difference image is calculated by calculating a difference value between the current frame image and the previous frame image of the deteriorated image, converting the difference value into an absolute value, and outputting the difference image. Steps,
Using the difference value of each pixel of the difference image obtained in the difference image calculation step as a class value, the frequency distribution calculation step of calculating the frequency distribution of the difference value of the difference image for the unit screen, and the frequency obtained in the frequency distribution calculation step. In order to divide the distribution into two distributions, a threshold calculation step of calculating a specific class value as a threshold, a pixel of a difference image having a lower limit of the threshold obtained in the threshold calculation step as a first value, and an upper limit of the threshold With the pixel of the difference absolute value image as the second value
A binarization step of replacing the binarized image with a binarized image; a block division step of dividing the binarized image obtained in the binarized step into blocks having a predetermined number of pixels to obtain a plurality of binarized image blocks; For each binarized image block obtained in the block division step, a counting step of counting the number of pixels of the first value, and a predetermined number of pixels of the first value among the blocks obtained in the counting step A binarized image block whose value is less than or equal to the distortion block is discriminated as a distortion block, and each pixel in the distortion block obtained in the distortion block discrimination step is smoothed to obtain an image with reduced noise. And a conversion step.

The invention of claim 6 of the present application is an image quality improving method for improving the distortion generated in the flat portion of the luminance near the contour in the vicinity of the contour of the degraded image in which the original image composed of a plurality of pixel data is degraded. Therefore, the degraded image is smoothed with a Gaussian distribution function having a predetermined standard deviation, and the smoothed image is secondarily differentiated to calculate a second derivative image.
Detecting a pixel that is an outline from the secondary differential image calculating step and the secondary differential image obtained in the secondary differential image calculating step,
A contour pixel detection step of obtaining a contour pixel; a binarization step of replacing the contour pixel obtained in the contour pixel detection step with a first value and other pixels with a second value to a binarized image; The binarized image obtained in the binarization step is divided into blocks each having a predetermined number of pixels to obtain a plurality of binarized image blocks, and each binarized image obtained in the block division step. For a block, a counting step of counting the number of pixels having a first value and a binarized image block in which the number of pixels having a first value is equal to or less than a predetermined value among the blocks obtained in the counting step is distorted. A distortion block determination step of determining a block,
A smoothing step of smoothing each pixel in the distortion block obtained in the distortion block determination step to obtain an image with reduced noise.

[0013]

The mosquito noise generated in the flat region near the strong contour in the image where the luminance change is large is visually noticeable.
On the other hand, the mosquito noise generated in the vicinity of the texture contour in which the luminance change is not so large and the luminance change periodically repeats is difficult to distinguish from the texture contour and is visually inconspicuous. When the mosquito noise generated in the flat area near the strong contour is reduced, the image quality is greatly improved. In many cases, the side effect of image quality deterioration occurs. Therefore, rather than reducing the mosquito noise generated near all contours of the image, noise reduction processing is avoided in the texture contour portion, and only the noise in the flat portion near the strong contour where mosquito noise is most noticeable is reduced. This is more effective in improving the image quality.

For this reason, according to the first and second aspects of the present invention, the deteriorated image in which the original image is deteriorated is divided into blocks by a predetermined number of pixels, and the in-block pixel data of the deteriorated image block is divided. Calculate the variation of. Next, in order to divide the block frequency distribution of the deteriorated image block with the variation value as a class value into three distributions, the first threshold and the first
A second threshold value that is smaller than the threshold value is calculated. First
The deteriorated image block whose lower limit is the threshold of is extracted as the first block including many strong contours with large luminance changes,
The deteriorated image block whose upper limit is the threshold of is extracted as the second block including many flat images. Further, the enlarged block of the first block is extracted, and the second block overlapping the enlarged block and the region is discriminated as a distorted block containing mosquito noise in the flat image near the strong contour. By smoothing only the finally obtained distorted block, the first threshold value that is likely to include the texture contour is set as the upper limit and the second threshold value is set.
The mosquito noise generated only in the flat part near the contour is reduced by avoiding the deteriorated image block whose lower limit is the threshold value of.

According to the third and fifth aspects of the present invention,
The deteriorated image in which the original image has been deteriorated is filtered by a secondary differential filter, and the absolute value of the value is taken to calculate a secondary differential image corresponding to the contour image. Next, the pixel frequency distribution of the secondary differential image with the secondary differential value of each pixel of the secondary differential image as a class value is obtained, and a threshold value for dividing the pixel frequency distribution into two distributions is calculated. 2 with a large change in brightness with the threshold being the lower limit
The binarized image is calculated with the pixel of the secondary differential image as the first value and the pixel of the secondary differential image with a small change in brightness with the threshold value as the upper limit as the second value. The binarized image is divided into blocks each having a predetermined number of pixels to calculate the binarized image block. The number of pixels of the first value is counted for each block of the binarized image block, and the number of pixels of the first value in the block is equal to or less than a predetermined number, and a binary value that does not form a connected contour or texture contour The digitized image block is discriminated as a distortion block. By smoothing the finally obtained distorted block, texture contours are avoided, and mosquito noise generated only in the flat portion near the contours is reduced.

Further, according to the inventions of claims 4 and 5 of the present application,
A difference image is calculated by absoluteizing the difference between the current frame image and the previous frame image of the deteriorated moving image in which the original image has deteriorated. Next, the pixel frequency distribution of the difference image with the difference value of each pixel of the difference image as the class value is obtained, and a threshold value for dividing the pixel frequency distribution into two distributions is calculated. A binarized image is calculated with a pixel having a difference value whose lower limit is a threshold value as a first value and a pixel having a difference value whose upper limit is a threshold value as a second value. The binarized image is divided into blocks each having a predetermined number of pixels to calculate the binarized image block. First for each block of the binarized image block
The number of pixels having the value of is counted, and the binarized image block in which the first value pixel number in the block is a predetermined number or less and does not include the difference pixel due to the movement of the connected contour or the texture contour is determined as the distorted block. By smoothing the finally obtained distorted block, texture contours are avoided, and mosquito noise generated only in the flat portion near the contours is reduced.

Further, according to the invention of claim 6 of the present application, the deteriorated image in which the original image is deteriorated is first smoothed by a Gaussian distribution function having a predetermined standard deviation and then second-order differentiated to detect a contour pixel. Next, the binarized image is calculated with the contour pixel as the first value and the other pixels as the second value. The binarized image is divided into blocks each having a predetermined number of pixels, and the binarized image block is calculated. The number of pixels of the first value is counted for each block of the binarized image block, and the number of pixels of the first value in the block is equal to or less than a predetermined number, so that a connected contour or texture contour is not formed. The binarized image block is identified as a distorted block. By smoothing the finally obtained distorted block, texture contours are avoided, and mosquito noise generated only in the flat portion near the contours is reduced.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An image quality improving method according to a first embodiment of the present invention will be described with reference to FIGS. 1, 2 and 3. FIG.
3 is a block diagram showing the arrangement of an image quality improving apparatus used in the image quality improving method of this embodiment. An image signal compressed by DCT or the like in an encoding device (not shown) is recorded on a recording medium or is output to a decoding device on the receiving side for transmission. The image quality improving apparatus shown in the figure is generally used when reproducing a recording medium or in a decoding apparatus on the receiving side.

The image signal decoded by the decoding device is often deteriorated as compared with the original image. An image signal of one frame including a deteriorated portion of such an image and composed of a plurality of blocks is input to the image memory 2 through the input terminal 1 of FIG. The image memory 2 is a memory that stores an input image signal in units of one frame, for example, and its output is given to the blocker 3. The blocker 3 has a function of a block division step for dividing one frame into a plurality of blocks. For example, an image signal of each block is set as a unit block including a region of 4 pixels in vertical and 4 pixels in horizontal. (Pixel signals) are sequentially output. The timing clock generator 4 is a timing clock signal generation circuit that controls the operations of the image memory 2, the blocker 3, and other circuit units described later.

The output of the blocker 3 is given to the intra-block deviation calculator 5 and the first delay memory 6. The intra-block deviation calculator 5 is a circuit that calculates the standard deviation σ of the luminance data of each pixel in the block (here, the data number n is 4 × 4 = 16) as a dispersion value, and has a function of a dispersion value calculation step. are doing. In-block deviation calculator 5
Outputs a total of m standard deviations when one frame is divided into m blocks. The delay memory 6 is a circuit that records the input image data of the block Bi (i = 1, 2, ..., M) for each block number i and adjusts the time with other circuit units. The output data of the intra-block deviation calculator 5 is given to the frequency distribution calculator 7 and the second delay memory 8.

The frequency distribution calculator 7 is a circuit having a function of a frequency distribution calculating step for classifying the input m standard deviations into class values having a predetermined class width and calculating the frequencies of blocks belonging to each class. Yes, the data is the threshold calculator 9
Given to. The delay memory 8 is the input block B
This is a circuit that holds the block number i and the standard deviation σi of i (i = 1 to m) and adjusts the time with other circuit units. The frequency distribution curves obtained by the frequency distribution calculator 7 are different curves depending on the type of image of each frame. That is, it is classified into a strong contour portion having a large luminance change, a flat portion having almost no luminance change, and a strong contour and a weak contour portion having a middle luminance change of the flat portion.

When the data of the frequency distribution regarding the standard deviation is input, the threshold calculator 9 selects two class values in order to divide the frequency distribution into three distributions. One class value is the first
, The other class value smaller than the first threshold value is set as the second threshold value. The threshold calculator 9 has a function of a threshold calculation step of calculating the first and second thresholds. The output of the threshold calculator 9 is given to the comparator 10.

The comparator 10 is a circuit for comparing the standard deviation of each block held in the delay memory 8 with the first threshold value and the second threshold value. The comparator 10 has the first block
When the block has a standard deviation equal to or more than the threshold of, it is determined to be a strong contour block. If the block has a standard deviation equal to or less than the second threshold, it is determined to be a flat block. Further, the comparator 10 determines a block having a standard deviation between the first threshold value and the second threshold value as a weak contour block.

The block distribution gate signal generator 11 is a circuit for generating a gate open signal based on the control signal of the timing clock generator 4 when the comparator 10 determines that the block is a strong contour block or a flat block. Here, the first gate opening signal is generated for the strong contour block, and the second gate opening signal is generated for the flat block, and these signals are output to the block distribution gate 12. When the first gate release signal is supplied, the block distribution gate 12 outputs the block number of the strong contour block to the strong contour block position bitmap calculator 13 shown in FIG. 2, and the second gate release signal is supplied. Then, the circuit outputs the block number of the flat block to the flat block position bitmap calculator 14.

The strong contour block position bitmap calculator 13 shown in FIG. 2 turns on only the bit at the position of the block number of the input strong contour block (black in FIG. 4A).
It is a circuit for calculating the strong contour block position bitmap by turning off the bits at the positions of the other block numbers. Flat block position bitmap calculator 14
Is a circuit for calculating a flat block position bitmap in which only the bits at the block number positions of the input flat blocks are turned on (black in FIG. 4B) and the bits at the other block number positions are turned off. is there. Here, the block distribution gate 12 and the strong contour block position bitmap calculator 13 have the function of the first extraction step of extracting the strong contour block. Further, the block distribution gate 12 and the flat block bitmap calculator 14 have a function of a second extraction step for extracting a flat block.

Enlarged block position bitmap generator 15
Turns on a bit in a predetermined range near the strong contour block input from the strong contour block position bitmap calculator 13 (hatched portion in FIG. 4C), and expands the block position bitmap (FIG. 4C). Is a circuit for generating a shaded area and a black portion) and has a function of an enlarged block calculating step. The distorted block position bitmap extractor 16 performs a logical product operation (AN) between the input flat block position bitmap and the input enlarged block position bitmap.
D) and perform the distortion block position bitmap (see FIG. 4).
This is a circuit for extracting the hatched portion (d).

The distortion block image extractor 17 shown in FIG.
On the basis of the control signal of the timing clock generator 4, the block number of the bit-on position of the distortion block bitmap input as shown in FIG. This is a circuit for replacing the block image with the number and extracting the distorted block image, and has a function of the distorted block determination step. The non-smoothed block image extractor 18 excludes the distorted block image extracted by the distorted block image extractor 17 from the block image supplied from the delay memory 6, and the non-smoothed block that does not perform the smoothing process described later. This is a circuit for extracting as an image.

The smoother 19 is a distortion block image extractor 1.
7 is a circuit for calculating a smoothed block image by filtering the distorted block image supplied from No. 7 with a predetermined smoothing filter and having a smoothing step function. The image synthesizer 20 synthesizes the non-smoothed block image supplied from the non-smoothed block image extractor 18 and the smoothed block image supplied from the smoother 19 to output an image with improved distortion. It is a circuit for outputting to 21.

The operation of the image quality improving apparatus of the first embodiment thus constructed will be described. First, in FIG. 1, an original image output from a signal source is supplied to an input terminal 1 with a luminance signal of a deteriorated image (hereinafter referred to as a deteriorated image signal) including a plurality of pixels deteriorated by DCT conversion, quantization, or the like. input. This deteriorated image signal is temporarily stored in the image memory 2,
The deteriorated image signal is supplied to the blocker 3 according to the control signal of the timing clock generator 4.

The blocker 3 divides the supplied 1-frame image signal into m blocks in accordance with the control signal from the timing clock generator 4 and sequentially outputs the blocks. Then, n pixel data in each block are supplied to the intra-block deviation calculator 5 and the delay memory 6. The intra-block deviation calculator 5 calculates the average value Ym of the brightness values Y of the n pixels in the block Bi and calculates the standard deviation σi as the brightness variation value. All standard deviations σi are calculated for m blocks, and those values are calculated as a frequency distribution calculator 7
And to the delay memory 8.

The frequency distribution calculator 7 sorts all the standard deviations σ into a predetermined class value and obtains the frequency distribution of the standard deviation σ. The threshold calculator 9 inputs the data of the frequency distribution,
And the second threshold value are calculated. The first threshold value and the second threshold value are calculated, for example, by dividing the frequency distribution into three distributions, that is, three distributions that maximize the variance among the three distributions. (D), vol.J6
An example of the discrimination criterion method is introduced in 3-D, pp.349-356, 1980. The first threshold value and the second threshold value calculated here are output to the comparator 10.

On the other hand, the delay memory 8 accumulates all the block numbers i in one frame and the standard deviation σi of the blocks during the calculation process of the frequency distribution calculation and the calculation of the first threshold value and the second threshold value. After the arithmetic processing, the data is compared by the comparator 10.
And the block distribution gate 12. The comparator 10 compares the supplied standard deviation σi of each block Bi with the first threshold value and the second threshold value. Then, in the case of standard deviation having the first threshold as the lower limit, a signal indicating a strong contour block is output, and in the case of standard deviation having the second threshold as the upper limit, a signal indicating a flat block is output. When the standard deviation is such that the first threshold value is the upper limit and the second threshold value is the lower limit, signals indicating weak contours are respectively supplied to the block distribution gate signal generator 1
Output to 1.

The block distribution gate signal generator 11 generates a gate open signal according to the control signal of the timing clock generator 4 only when a signal indicating a strong contour block and a signal indicating a flat block are supplied. That is, when a signal indicating a strong contour block is supplied, a first gate opening signal is generated, and when a signal indicating a flat block is supplied, a second gate opening signal is generated and output to the block distribution gate 12. To do.

The block distribution gate 12 opens the gate of the m blocks stored in the delay memory 8 only when the first gate opening signal or the second gate opening signal is input. That is, when the first gate opening signal is input, the block number of the strong contour block is output to the strong contour block position bitmap calculator 13 in FIG. 2, and when the second gate opening signal is input, it is flattened. The block number of the block is output to the flat block position bitmap calculator 14.

Strong Contour Block Position Bitmap Calculator 1
As shown in FIG. 4A, 3 calculates a strong contour block position bitmap in which bits are turned on only for the position of the block number of the supplied strong contour block. Then, the enlarged block position bitmap generator 15 turns on the bit in a predetermined range in the vicinity of ON of the strong contour block position bitmap, and the enlarged block position bitmap of the strong contour block position bitmap is displayed as shown in FIG. 4C. To generate.

This enlarged block position bitmap is supplied to the distorted block position bitmap extractor 16. The flat block position bitmap calculator 14 calculates a flat block position bitmap in which bits are turned on only for the position of the block number of the supplied flat block, and outputs the flat block position bitmap to the distorted block position bitmap extractor 16. The distorted block position bitmap extractor 16 performs a logical product of the flat block position bitmap supplied from the flat block position bitmap calculator 14 and the expanded block position bitmap supplied from the expanded block position bitmap generator 15. Calculate. Then, a distorted block position bitmap in which only the bits whose on-neighborhood bits of the strong contour block position bitmap are on and the flat block bitmap shown in FIG. The distortion block position bitmap extracted here is as shown in FIG. 4D and is output to the distortion block image extractor 17 in FIG.

The distorted block image extractor 17 is responsive to the control signal from the timing clock generator 4 for delay memory 6.
Block images (B1, B2, B3.
.. of Bi, Bj, Bk, Bl, and Bm), only the block image (for example, Bj, Bk, Bl) having the bit-on block number of the input distortion block position bitmap is extracted to obtain the distortion block image . The distorted block image (Bj, Bk, Bl) extracted here is output to the non-smoothed block image extractor 18 and the smoother 19.

The non-smoothed block image extractor 18 is a non-smoothed block image which is the block image excluding the distorted block image supplied from the distorted block image extractor 17 out of the m block images stored in the delay memory 6. Block image (B
1, B2, B3 ... Bi, Bm) are extracted and output to the image synthesizer 20. The smoother 19 receives the distortion block image (Bj, Bj supplied from the distortion block image extractor 17).
k, Bl) is filtered by a smoothing filter, for example, a filter composed of 3 × 3 pixels as shown in Formula 1, and the distorted pixels are averaged by the brightness values of the peripheral pixels to reduce the distortion. bi, bk, bl) are calculated.

[Equation 1]

The smoothed block image calculated here is output to the image synthesizer 20. The image synthesizer 20 supplies the non-smoothed block images (B1, B2, B3 ... Bi, Bm) supplied from the non-smoothed block image extractor 18 and the smoothed block images (B1, B2, B3 ... Bi, Bm) supplied from the smoother 19 ( bj, b
k, bl) are combined and the distortion-improved images (B1, B2, B3
... Bi, bj, bk, bl, Bm) are output to the output terminal 21.

In the above embodiment, the block size is 4 pixels horizontally by 4 pixels vertically, but the size of the block of the image which is divided into blocks at the time of encoding and compression encoded (for example, 8 × 8 pixels). ). In addition, the standard deviation of the image data was used to calculate the deviation within the block.
A variance value may be used. Further, no post-treatment was intentionally performed on the weak contour block including texture and the like, but this is not limited to the case where the image pattern or the like is known and a specific filter for noise reduction or contour enhancement can be designed.

As described above, according to the image quality improving method of the first embodiment, the deteriorated image is block-by-block in units of a strong contour block including a contour having a large luminance change, a flat block having little luminance change, and a strong contour block and a flat block. It is possible to divide into three blocks, a weak contour block located in the middle of the block and a flat block in the vicinity of the strong contour block, and smooth only those blocks. Therefore, it is possible to efficiently extract a region where mosquito noise is generated in a flat region near the contour that is visually conspicuous and reduce the amount of calculation to reduce noise. at the same time,
The texture contour region which is visually unnoticeable and which rather deteriorates the image quality when the filtering process is performed can be left unprocessed.

Next, an image quality improving method according to the second embodiment of the present invention will be described with reference to FIGS. 5 and 6 are block diagrams showing the configuration of an image quality improving apparatus for realizing the image quality improving method in this embodiment. The same parts as those in the first embodiment are designated by the same names and detailed description thereof will be omitted.

An image signal compressed by DCT or the like by an encoding device (not shown) is recorded on a recording medium or is output to a decoding device on the receiving side for transmission. The deteriorated image signal including the deteriorated portion decoded by the decoding device is input terminal 31.
Is given to the image memory 32 via. Image memory 32
Is a memory that stores an input image in units of, for example, one frame, and this output is a secondary differentiation circuit 35 and a delay memory 3
Given to 4. The delay memory 34 is a circuit that holds the deteriorated image signal and adjusts the time with other circuit units.

The timing clock generator 33 is a control signal generating circuit for controlling the operation timing of the image memory 32, the blocker described later, and the gate signal generator.
The secondary differentiation circuit 35 uses a secondary differentiation filter for the degraded image signal,
For example, a Laplacian filter as shown in the following equation is a circuit that calculates a secondary differential signal indicating the rate of change in the luminance change of all r pixels in the image, and has the function of the secondary differential image calculation step. .

[Equation 2] The absolute value circuit 36 is a circuit for outputting the absolute value of the input r second-order differential signals.

The frequency distribution calculator 37 receives the input r 2
It is a circuit that classifies the secondary differential value into class values having a predetermined class width and calculates the frequency of pixels belonging to each class, and has a function of a frequency distribution calculation step. Then, these data are given to the threshold calculator 39. On the other hand, the delay memory 38 is a circuit that holds the input r pixel data values and adjusts the time with other circuit units. Frequency distribution calculator 3
The frequency distribution curve obtained in 7 is different depending on the type of image in each frame. This curve differs depending on the ratio of contour pixels such as contours having a large secondary differential value and flat pixels such as a background having a small secondary differential value. The threshold value calculator 39 is a circuit for calculating a threshold value (class value) for dividing the frequency distribution into two distributions when the data of the frequency distribution of the second derivative is input, and has a function of a threshold value calculation step. ing. The output of the threshold calculator 39 is the comparator 4
Given to 0.

The comparator 40 compares the secondary differential value of each pixel held in the delay memory 38 with a threshold value.
It is a circuit that discriminates a pixel having a secondary differential value having a lower limit of a threshold value as a contour pixel and a pixel having a secondary differential value having a threshold value of an upper limit as a flat pixel. The binarization circuit 41 sets the second-order differential value of each of the r pixels stored in the delay memory 38 to a first value when it is determined to be a contour pixel and a second value when it is determined to be a flat pixel. It is a circuit that substitutes a value and calculates a binarized image, and has a function of a binarization step. Based on the timing clock generator 33, the blocker 42 divides one input binarized image frame into m blocks, and each block Bi (i = 1, 2, ...).
It is a circuit for sequentially outputting the binarized signal of m) and has a function of a block division step. The blocker 43 divides the input degraded image signal into m blocks on the basis of the timing clock generator 33, and also outputs pixel data of each block Bi (i = 1, 2, ... M). It is a circuit that outputs sequentially.

The counter 44 of FIG. 6 is a circuit which counts and outputs the number of signals having the first value in the block, that is, the number of contour pixels Si, out of the input binarized signal of the block Bi. It has the function of. The threshold value memory 45 has a threshold value for determining whether or not the block includes a connected contour or a texture contour based on the number of contour pixels in the block, and supplies this threshold value to the comparator 46. The comparator 46 receives the supplied threshold value and each block Bi
Of the contour pixel number Si of the block distribution gate, and the block of the contour pixel number having the lower limit of the threshold value is a signal indicating the contour block, and the block of the contour pixel number of the upper limit value of the threshold value is the signal indicating the flat block. This is a circuit for outputting to the signal generator 47.

Based on the timing clock generator 33, the block distribution gate signal generator 47 outputs a first gate opening signal when a signal indicating a contour block is supplied, and is also supplied with a signal indicating a flat block. Then, the circuit outputs the second gate opening signal to the block distribution gate 48. When the first gate opening signal is input, the block distribution gate 48 outputs the in-block pixel data of the degraded image signal to the non-smoothed block image memory 50, and the second
When the gate opening signal is input, the circuit outputs the signal to the smoothed block image memory 49. Here, the comparator 46,
The block distribution gate signal generator 47 and the block distribution gate 48 have a function of a smoothing block judgment step for judging and selecting a distortion block.

The smoothed block image memory 49 is a memory for storing block pixel data of a deteriorated image having the same block number as the block number of the contour block. The non-smoothed block image memory 50 is a memory that stores block pixel data of a deteriorated image having the same block number as the block number of the flat block. Smoother 51, image synthesizer 5
2 and the output terminal 21 are respectively the smoothing device 1 of the first embodiment.
9. It has the same function as the image synthesizer 20.

The operation of the image quality improving apparatus of the second embodiment having the above configuration will be described. First, the input terminal 3 of FIG.
1, the deteriorated image signal is input in frame units. The deteriorated image signal is temporarily stored in the image memory 32, and the deteriorated image signal is stored in accordance with the control signal of the timing clock generator 33.
It is supplied to the next differentiating circuit 35. Further, the deteriorated image signal stored in the image memory 32 is input to the delay memory 34, and the delay memory 34 time-adjusts the deteriorated image signal with other circuit parts and outputs the deteriorated image signal to the blocker 43. Supply.

The secondary differentiating circuit 35 calculates a secondary differential signal for all r image signals by filtering the deteriorated image signal. The absolute value circuit 36 converts the secondary differential signal into an absolute value, and converts the signal into a frequency distribution calculator 37 and a delay memory 38.
Output to. The frequency distribution calculator 37 sorts the supplied secondary differential value (absolute value) into a predetermined class width, and calculates the frequency distribution of the secondary differential value. The threshold value calculator 39 inputs the frequency distribution data and calculates a threshold value (class value). The method of calculating the threshold value here is to divide the frequency distribution into two distributions, that is, two distributions in which the distribution between the two distributions is maximum, and details thereof will be described in the threshold value calculator 9 of the first embodiment. Same as the case. The threshold value calculated here is output to the comparator 40.

On the other hand, the delay memory 38 accumulates the secondary differential value of the pixel in one frame during the calculation of the frequency distribution calculation and the threshold calculation, and after the calculation processing, the data is used as the comparator 40 and the binarization circuit 41. Supply to. The comparator 40 compares the supplied second-order differential value of each pixel with a threshold value, and outputs a signal indicating a contour pixel when the second-order differential value has a threshold value as a lower limit, and a second-order value with a threshold value as an upper limit. In the case of the differential value, a signal indicating a flat pixel is output to the binarization circuit 41. The binarization circuit 41 binarizes the secondary differential values of the pixels sequentially supplied from the delay memory 38 by the signal input from the comparator 40,
A binary image is calculated. In this case, when the signal indicating the contour pixel is input from the comparator 40, the secondary differential value of the pixel is replaced with the second value. The binarized image calculated here is output to the blocker 42.

The blocker 42 divides an area consisting of a total of 16 pixels, for example, 4 vertical pixels × 4 horizontal pixels, into a block having 16 binarized pixel data, and binarizes each block. The pixel data is sequentially output to the counter 44 of FIG. The counter 44 counts the number of signals of the first value among the binarized signals in the block, that is, the number Si of contour pixels, and sequentially outputs the count to another comparator 46. The threshold memory 45 holds a predetermined threshold value and outputs the value to the comparator 46. The comparator 46 compares the supplied contour pixel number Si of each block Bi with the threshold value supplied from the threshold value memory 45, and when the threshold pixel number is the lower limit, outputs a signal indicating the contour block, If the number of contour pixels is the upper limit of the threshold value, a signal indicating a flat block is output and given to the block distribution gate signal generator 47.

The block distribution gate signal generator 47 operates based on the control signal of the timing clock generator 33, and when the signal indicating the contour block is supplied, outputs the first gate opening signal. When a signal indicating a flat block is supplied, a second gate opening signal is output and given to the block distribution gate 48. The block distribution gate 48 outputs the in-block pixel data of the deteriorated image signal supplied from the blocker 43 to the non-smoothed block image memory 50 when the first gate opening signal is input, and the second gate opening is performed. When the signal is input, it is output to the smoothed block image memory 49.

The smoothed block image memory 49 stores the block pixel data of the supplied deteriorated image having the same block number as the block number of the contour block, and supplies it to the smoother 51 having a smoothing step function. The non-smoothed block image memory 50 stores the block pixel data of the supplied degraded block having the same block number as the block number of the flat block, and supplies the block pixel data to the image synthesizer 52. Here, the smoothing device 51 and the image synthesizing device 52 perform the same operations as the smoothing device 19 and the image synthesizing device 20 of the first embodiment, respectively, and output an image with improved image quality from the output terminal 53.

As described above, according to the second embodiment, the deteriorated image is first discriminated from the contour image having a large luminance change according to the magnitude of the second derivative of the pixel. Next, the binarized image binarized with the contour image and the other pixels is divided into blocks, and blocks in which the number of contour pixels occupying in the block is a predetermined number or less are extracted as blocks containing mosquito noise. And smooth it. In this case, a block including a predetermined number or more of contour images in the block can be regarded as a main contour or texture contour in which the contour images are connected, and smoothing of these blocks can be avoided.

Next, an image quality improving method according to the third embodiment of the present invention will be described with reference to FIGS. Figure 7
FIG. 8 is a block diagram showing the arrangement of an image quality improving apparatus for implementing the image quality improving method in this embodiment. Figure 7
Further, in FIG. 8, circuits having the same functions as those in the second embodiment are given the same names, and detailed description thereof will be omitted.

An image signal compressed by DCT or the like in an encoding device (not shown) is recorded on a recording medium or is output to a decoding device on the receiving side for transmission. At the input terminal 61 of FIG. 7, a signal of a deteriorated image composed of a plurality of frames including a deteriorated portion decoded by the decoding device (a deteriorated image signal)
Of these, the image signal of the current frame (n frames) is input. Further, the image signal of the previous frame (n-1 frame), which is one frame before the current frame, is input to the input terminal 62. The image memory 63 is a memory for storing the image of the current frame, and this image is given to the difference circuit 67 and the delay memory 66. The image memory 64 is a memory for storing the image of the previous frame, and this image is the difference circuit 67.
Given to.

The timing clock generator 65 is a control signal generating circuit for controlling the operation timing of the image memory 63, the image memory 64, the blocker 74, the blocker 75 and the block distribution gate signal generator 79. The delay memory 66 holds the deteriorated image signal of the current frame,
It is a circuit for adjusting the time with other circuit parts.

The difference circuit 67 is a circuit which calculates the difference between the deteriorated image of the current frame and the deteriorated image of the previous frame at corresponding pixel positions and generates a difference image, and has a function of a difference image calculation step. There is. The absolute value circuit 68 is a circuit that outputs a difference image signal obtained by converting the positive value and the negative value of the input difference image into absolute values. The frequency distribution calculator 69
It is a circuit that classifies the difference value of each pixel in the input difference image into a class value having a predetermined class width and calculates the frequency of pixels belonging to each class, and has a function of a frequency distribution calculation step. . The delay memory 70 is a circuit that holds the pixel data value of the input difference image and adjusts the time with other circuit units. The threshold value calculator 71 is a circuit for calculating a threshold value (class value) for dividing the frequency distribution into two distributions when the data of the frequency distribution of the difference value is input, and has a function of a threshold value calculation step. .

The comparator 72 compares the difference value of each pixel held in the delay memory 70 with a threshold value. The pixel having the difference value whose lower limit is the threshold value is determined as the contour pixel and the threshold value is set as the upper limit. This is a circuit for discriminating a pixel having a difference value as a flat pixel. Binarization circuit 7 having a function of a binarization step
3 is a binary value in which the difference value of each pixel stored in the delay memory 70 is replaced with a first value when it is determined to be a contour pixel and a second value when it is determined to be a flat pixel. It is a circuit for calculating a digitized image. The blocker 74 of FIG.
75, a counter 76, a threshold memory 77, a comparator 78, a block distribution gate signal generator 79, a block distribution gate 80, a smoothed block image memory 81, a non-smoothed block memory 82, and a smoother 83 shown in FIG. The configurations and functions of the image synthesizer 84 and the output terminal 85 are the same as those of the second embodiment, and the description thereof will be omitted.

The operation of the image quality improving apparatus of the third embodiment having the above-described structure will be described, especially regarding the difference from the second embodiment. First, the current frame (n frames) of the degraded image signal is input to the input terminal 61 of FIG. 7, and the previous frame (n−1 frame) of the degraded image signal is input to the input terminal 62. The degraded image signal of the current frame is temporarily stored in the image memory 63, and the timing clock generator 6
The differential circuit 67 and the delay memory 6 according to the control signal 5
6. The deteriorated image signal of the previous frame is temporarily stored in the previous image memory 64 and is supplied to the difference circuit 67 according to the control signal of the timing clock generator 65. The delay memory 66 time-adjusts the deteriorated signal with other circuit units, and supplies the deteriorated image signal of the current frame to the blocker 75.

The difference circuit 67 calculates a difference signal between the supplied deteriorated image signal of the current frame and the supplied deteriorated image signal of the previous frame, and supplies it to the absolute value circuit 68. Absolute value circuit 6
Reference numeral 8 converts the value of the input difference signal into an absolute value, and outputs the value to the frequency distribution calculator 69 and the delay memory 70.
The frequency distribution calculator 69 sorts the difference values (absolute values) of the supplied difference signals into predetermined class widths, and calculates the frequency distribution of the difference values. The frequency distribution calculated here is the threshold value calculator 7
1 is supplied. The threshold calculator 71 calculates a threshold (class value) from the supplied frequency distribution data. The method of calculating the threshold is to divide the frequency distribution into two distributions, that is, two distributions that maximize the variance between the two distributions, and the details thereof are the same as the discrimination criterion method of the second embodiment. The threshold value calculated here is output to the comparator 72.

On the other hand, the delay memory 70 accumulates the pixel difference values in one frame during the calculation of the frequency distribution calculation and the threshold calculation, and supplies the data to the comparator 72 and the binarization circuit 73 after the calculation processing. . The comparator 72 compares the supplied difference value of each pixel with a threshold value, and in the case of a pixel having a difference value whose lower limit is the threshold value, outputs a signal indicating a contour pixel, and outputs the signal of the difference value whose upper limit is the threshold value. In this case, the signal indicating the flat pixel is output to the binarization circuit 73. The binarization circuit 73 compares the pixel difference values sequentially supplied from the delay memory 70 with the comparator 7.
It is binarized by a signal input from 2 and a binarized image is calculated. In this case, when the signal indicating the contour pixel is input from the comparator 72, the difference value of the pixel at that time is replaced with the first value, and when the signal indicating the flat pixel is input, the pixel at that time is input. The difference value of is replaced with the second value. The binarized image calculated here is output to the blocker 74. The operation of the device from the blocker 74 of FIG. 7 to the output terminal 85 of FIG. 8 is the same as the operation of the blocker 42 of the second embodiment to the output terminal 53, and a description thereof will be omitted.

As described above, according to the third embodiment, first, a pixel whose luminance greatly changes between frames is discriminated as a contour image according to the magnitude of each image difference value of the inter-frame difference image of the deteriorated image. Next, the binarized image binarized by the contour image and the other pixels is divided into blocks, and blocks in which the number of contour pixels occupying in the block is equal to or less than a predetermined number are extracted as blocks including mosquito noise. Smooth it. In this case, a block including a predetermined number or more of contour images in the block can be regarded as a main contour or texture contour in which the contour images are connected, and smoothing of these blocks can be avoided.

Next, an image quality improving method according to the fourth embodiment of the present invention will be described with reference to FIGS. 9 and 10. 9 and 10 are block diagrams showing the configuration of an image quality improving apparatus for realizing the image quality improving method in this embodiment. In FIGS. 9 and 10, circuits having the same functions as those in the second and third embodiments have the same names, and detailed description thereof will be omitted.

An image signal compressed by DCT or the like in an encoding device (not shown) is recorded on a recording medium or output to a decoding device on the receiving side for transmission. The deteriorated image signal including the deteriorated portion decoded by the decoding device is input to the input terminal 91 of FIG. 9 in units of one frame. The image memory 92 is a memory that stores an image for one frame,
This image is supplied to the Laplacian Gaussian circuit 95 and the delay memory 94. Timing clock generator 9
Reference numeral 3 denotes a control signal generation circuit for controlling the operation timing of the image memory 92, the blocker 98, the blocker 99, and the block distribution gate signal generator 103. The delay memory 94 is a circuit that holds a degraded image signal for one frame and adjusts the time with other circuit units.

The Laplacian Gaussian circuit 95 is a circuit equivalent to a series connection of the smoothing circuit 110 and the secondary differentiating circuit 111, and has the function of the secondary differential image calculating step. Laplacian-Gaussian circuit 95 is multiplied by the Laplacian operator ▽ ² to a Gaussian distribution function G of the standard deviation sigma (x, y), performs a calculation indicated by the following equation ^{▽ 2 · G (x, y} ).

(Equation 3) The smoothing circuit 110 performs smoothing with a Gaussian distribution function,
The secondary differentiating circuit 111 performs the secondary differential of the smoothed signal.

The Laplacian-Gaussian filter (Laplacian-Gaussian circuit 95) represented by ▽ ² · G (x, y) is an operator in the rotationally symmetric Mexican hat shown in FIGS. 11 (a) and 11 (b). , The width ω of the area of the central convex portion is given by the following equation.

[Equation 4] Further, the range λ covered by the filter is given by the following equation.

(Equation 5)

The Laplacian Gaussian filter is a kind of differential filter, and by using this filter, FIG.
An image signal I (x, y) having a luminance change as shown in (c)
Is processed, the output characteristic has peaks (+) and valleys (-) as shown in FIG. Also, the Laplacian Gaussian filter can change the filter width by the standard deviation, and can change the degree of smoothing. Here, such an output can be obtained when there is a change in luminance more than mosquito noise.

The contour detector 96 outputs a signal (zero crossing signal) of the secondary differential signal supplied from the Laplacian-Gaussian circuit 95, which changes from positive or negative to negative or positive, respectively, to the contour pixel. It is a circuit that detects a signal and outputs it to the binarization circuit 98 together with other non-zero crossing signals, and has a function of a contour pixel detection step. Two
The binarizing circuit 97 replaces the signal supplied from the contour detector 96 with a first value when the signal is a zero-crossing signal and replaces it with a second value when the signal is a non-zero-crossing signal to calculate a binarized image. And has a function of a binarization step. Since the configuration and function from the blocker 98 in FIG. 9 to the output terminal 109 in FIG. 10 are the same as the circuit from the blocker 42 to the output terminal 53 in the second embodiment, the description of the configuration is omitted. To do.

The operation of the image quality improving apparatus having the above-described structure will be described with particular reference to differences from the second and third embodiments. First, the degraded image signal is input to the input terminal 91 of FIG. The deteriorated image signal is temporarily stored in the image memory 9
2 and is supplied to the Laplacian Gaussian circuit 95 and the delay memory 94 according to the control signal of the timing clock generator 93. The delay memory 94 adjusts the time of the deteriorated image signal in relation to other circuit parts, and supplies the deteriorated image signal to the blocker 99.

The Laplacian Gaussian circuit 95 filters the input deteriorated image signal with a Laplacian Gaussian filter having a predetermined standard deviation. This filter processing is equivalent to the smoothing processing by the Gaussian distribution function of the smoothing circuit 110 and the filter processing by the Laplacian filter of the second derivative circuit 111, and outputs the second derivative signal of the deteriorated image signal. . This second derivative signal is output to the contour detector 96. The contour detector 96 detects the signal of the contour pixel, which is a zero-crossing signal whose signal value changes from positive to negative or from negative to positive, from the input second-order differential signal, and the binarization circuit 97 together with the non-zero-crossing signal. Output to.
The binarization circuit 97 replaces the signal input from the contour detector 96 with the first value when the signal is a zero-crossing signal, and replaces it with the second value when the signal is a non-zero-crossing signal to calculate a binary image. . The binarized image calculated here is supplied to the blocker 98. The operation from the blocker 98 to the output terminal 109 is the same as that in the second embodiment, and the description thereof will be omitted.

As described above, according to the fourth embodiment, the Laplacian-Gaussian filter having a predetermined standard deviation is first used to convert the deteriorated moving image into a second-order differential image having a luminance change equal to or higher than mosquito noise, and the second-order differential image is obtained. The contour pixel is detected from the differential image. Next, the binarized image binarized by the contour image and the non-contour pixels is divided into blocks, and blocks in which the number of contour pixels occupying in the block is a predetermined number or less are extracted as blocks including mosquito noise and smoothed. Turn into.
In this case, a block including a predetermined number or more of contour images in the block can be regarded as a main contour or texture contour in which the contour images are connected, and smoothing of these blocks can be avoided.

As described above, in each of the first to fourth embodiments, the smoothing process is avoided in the image portion including the texture contour, and only the mosquito noise image portion generated in the flat area near the strong contour is smoothed. I am trying to do it. Therefore, it is possible to prevent blurring of the texture contour, which is a conventional problem. However, since each embodiment has a different configuration, it has the following individual characteristics.

In the first embodiment, the blocks are classified into three groups according to the size of the variation value of the pixel data in the blocks, and the noise blocks are extracted from the adjacency relationship between the blocks. Therefore, the processing can be consistently performed in block units, so that the amount of calculation can be reduced as compared with the other three embodiments. However, since the blocks are classified into three, it is necessary to calculate two thresholds.

In the second embodiment, a contour pixel having a large change in luminance is discriminated by the magnitude of the second derivative value of the pixel, and a noise block is extracted from the number of contour pixels included in a fixed block. . Therefore, if the secondary differential image is calculated, the processing can be performed with one threshold value calculation.

In the third embodiment, an outline pixel having a large brightness change is discriminated by the size of the difference value of the pixels of the inter-frame difference image, and the noise block is determined from the number of outline pixels included in a certain block. To extract. Therefore, if the difference image is calculated, the processing can be performed with one threshold value calculation. Note that this method can be applied only to a deteriorated moving image, but since mosquito noise is noise that is originally more annoying due to noise increase / decrease between frames, it has a great improvement effect in a moving image.

In the fourth embodiment, a contour pixel having a large change in luminance is obtained by detecting a contour pixel from a quadratic differential image accompanied by a predetermined smoothing, and the contour pixel is included in a certain block. Extract noise blocks. For this reason, the calculation of the secondary differential image and the detection of the contour pixel are required, but the threshold value need not be calculated.

[0080]

As described above, in any of the inventions of the present application, in reducing the distortion generated in the vicinity of the contour, the distortion reducing effect is small and the distortion in the fine texture contour portion is likely to give the side effect of blurring or distortion. Reduction can be avoided. Therefore, the image quality can be improved by reducing only the distortion generated in the flat image region in the vicinity of the contour portion where the distortion is more noticeable and in which the luminance change is large.

Particularly, according to the inventions of claims 1 and 2, the deteriorated image is divided into flat blocks, strong contour blocks and weak contour blocks in block units. Mosquito noise can be effectively reduced by extracting a flat block near the strong contour block and smoothing the image. Further, the image quality of the texture contour area does not deteriorate.

According to the third and sixth aspects of the invention, the contour image is discriminated by the magnitude of the secondary differential value of the deteriorated image. By doing this, it is possible to improve the image quality without deteriorating the image quality of the texture contour region by extracting the block in which the number of contour pixels occupying in the block is a predetermined number or less as a block containing mosquito noise and smoothing the image. You can

Further, according to the inventions of claims 4 and 5, the contour image is discriminated by the magnitude of the difference value between the frames of the deteriorated image. Therefore, the image quality can be effectively improved even with respect to mosquito noise generated in a moving image.

[Brief description of drawings]

FIG. 1 is a block diagram of an image quality improving apparatus (No. 1) that implements an image quality improving method according to a first embodiment of the present invention.

FIG. 2 is a block diagram of an image quality improving device (No. 2) for realizing the image quality improving method of the first embodiment.

FIG. 3 is a block diagram of an image quality improving device (No. 3) for implementing the image quality improving method of the first embodiment.

FIG. 4A is an operation explanatory diagram of the strong contour block position bitmap calculator in the first embodiment. FIG. 7B is an explanatory diagram of the operation of the flat block position bitmap calculator in the first embodiment. FIG. 7C is an explanatory diagram of the operation of the enlarged block position bitmap calculator in the first embodiment.
FIG. 6D is an operation explanatory diagram of the distortion block position bitmap calculator in the first embodiment.

FIG. 5 is a block diagram of an image quality improving device (No. 1) for realizing the image quality improving method according to the second embodiment of the present invention.

FIG. 6 is a block diagram of an image quality improving device (No. 2) for realizing the image quality improving method of the second embodiment.

FIG. 7 is a block diagram of an image quality improving device (No. 1) for realizing the image quality improving method according to the third embodiment of the present invention.

FIG. 8 is a block diagram of an image quality improving device (No. 2) that realizes the image quality improving method of the second embodiment.

FIG. 9 is a block diagram of an image quality improving device (No. 1) for realizing the image quality improving method according to the fourth embodiment of the present invention.

FIG. 10 is a block diagram of an image quality improving device (No. 2) for implementing the image quality improving method of the fourth embodiment.

FIG. 11 is an operation explanatory diagram of the Laplacian Gaussian circuit according to the fourth embodiment of the present invention.

[Explanation of symbols]

1, 31, 61, 62, 91 Input terminal 2, 32, 63, 64, 92 Image memory 3, 42, 43, 74, 75, 98, 99 Blocker 4, 33, 65, 93 Timing clock generator 5 In-block deviation calculator 6,8,34,38,66,70,94 Delay memory 7,37,69 Frequency distribution calculator 9,39,71 Threshold calculator 10,40,46,72,78,102 Comparison Device 11, 47, 79, 103 Block distribution gate signal generator 12, 48, 80, 104 Block distribution gate 13 Strong contour block position bitmap calculator 14 Flat block position bitmap calculator 15 Enlarged block position bitmap calculator 16 Distorted block position bitmap calculator 17 Distorted block image extractor 18 Unsmoothed block image extractor 19, 51, 83,107 Smoothing device 20,52,84,108 Image synthesizer 21,53,85,109 Output terminal 35,111 Second derivative circuit 36,68 Absolute value circuit 41,73,97 Binarization circuit 44,76 , 100 counter 45, 77, 101 threshold memory 49, 81, 105 smoothed block image memory 50, 82, 106 non-smoothed block image memory 67 difference circuit 95 Laplacian Gaussian circuit 96 contour detector 110 smoothing circuit

Claims (6)

[Claims] 1. An image quality improving method for improving image distortion generated in a flat portion of luminance near the contour in the vicinity of the contour of the deteriorated image in which an original image composed of a plurality of pixel data has deteriorated. A block division step of dividing an image into blocks having a predetermined number of pixels to obtain a plurality of deteriorated image blocks, and for each deteriorated image block obtained in the block division step, a variation of pixel data in the block is calculated. A dispersion value calculating step, a dispersion value obtained in the dispersion value calculating step as a class value, and a frequency distribution calculating step of calculating a frequency distribution of the dispersion values of the deteriorated image block for a unit screen; and the frequency distribution calculating step. The first threshold, which is two class values for dividing the frequency distribution obtained in 3 into 3 according to the strength of the image contour, and A threshold value calculation step of calculating a second threshold value that is smaller than the first threshold value, and the deteriorated image block whose lower limit is the first threshold value obtained in the threshold value calculation step is extracted as a first block. A first extraction step; a second extraction step of extracting, as a second block, the deteriorated image block whose upper limit is the second threshold value obtained in the threshold value calculation step; and a second extraction step obtained in the first extraction step. The enlarged block calculating step of enlarging the obtained first block to a predetermined size and calculating the enlarged block, and the second block obtained in the second extracting step are obtained in the enlarged block calculating step. A distortion block determination step of determining a block whose area overlaps the enlarged block obtained as a distortion block, and the distortion obtained in the distortion block determination step. Image quality improving method characterized by comprising a smoothing step of obtaining a smoothed reduced image noise for each pixel in the block. 2. The threshold value calculating step is a step of determining the first threshold value and the second threshold value so as to maximize the inter-distribution variance of the three distributions. Described image quality improvement method. 3. An image quality improving method for improving distortion generated in a flat portion of luminance near the contour in the vicinity of the contour of the degraded image in which an original image composed of a plurality of pixel data has deteriorated. To a second-order differential image calculation step for calculating a second-order differential image by converting the value obtained by the second-order differential image into an absolute value, and 2 for each pixel of the second-order differential image obtained in the second-order differential image calculation step. A frequency distribution calculation step of calculating the frequency distribution of the secondary differential values for a unit screen using the secondary differential value as a class value, and two frequency distributions obtained by the frequency distribution calculation step depending on the strength of the image contour. A threshold value calculation step of calculating a specific class value as a threshold value, and a pixel of the secondary differential image having a lower limit of the threshold value obtained in the threshold value calculation step as a first value, A binarization step of replacing the unit screen with a binarized image using pixels of the secondary differential image whose upper limit is the threshold value as a second value, and the binarized image obtained in the binarization step. Is divided into blocks each having a predetermined number of pixels to obtain a plurality of binarized image blocks, and for each binarized image block obtained in the block dividing step, the first value of A counting step of counting the number of pixels, and a smoothing block that determines, as a distorted block, the binarized image block in which the number of pixels of the first value is equal to or less than a predetermined number among the blocks obtained in the counting step. And a smoothing step of smoothing each pixel in the distortion block obtained in the smoothing block determination step to obtain an image with reduced noise. Image quality improvement method according to symptoms. 4. An image quality improving method for improving image distortion generated in a flat portion of brightness near the contour in the vicinity of the contour of the deteriorated image in which the original image composed of a plurality of pixel data has deteriorated. When the image includes a moving image composed of a plurality of frames, a difference image calculation step of calculating a difference value between the current frame image and the previous frame image of the deteriorated image and outputting the difference image by converting the difference value into an absolute value, Using the difference value of each pixel of the difference image obtained in the difference image calculation step as a class value, the frequency distribution calculation step of calculating the frequency distribution of the difference value of the difference image for the unit screen, and the frequency distribution calculation step In order to divide the obtained frequency distribution into two distributions, a threshold value calculation step of calculating a specific class value as a threshold value, and the threshold value obtained in the threshold value calculation step as a lower limit A binarization step of replacing the unit screen with a binarized image by setting a pixel of the difference image as a first value and a pixel of the difference absolute value image whose upper limit is the threshold value by a second value; The binarized image obtained in the binarizing step is divided into blocks having a predetermined number of pixels to obtain a plurality of binarized image blocks, and each binarized image obtained in the block dividing step. A counting step of counting the number of pixels having the first value with respect to a block, and the binarization in which the number of pixels having the first value is equal to or less than a predetermined value among the blocks obtained in the counting step. A distortion block determination step of determining an image block as a distortion block, and smoothing each pixel in the distortion block obtained in the distortion block determination step to obtain an image with reduced noise Image quality improving method characterized by comprising the smoothing step. 5. The threshold value calculating step calculates the inter-distribution variance of the two distributions when calculating the frequency distribution obtained in the frequency distribution calculating step as a threshold value which divides into two distributions according to the strength of the image contour. The image quality improving method according to claim 3 or 4, wherein the determining step is performed so as to maximize the image quality. 6. An image quality improving method for improving distortion generated in a flat portion of luminance near the contour in the vicinity of the contour of the degraded image in which an original image composed of a plurality of pixel data has deteriorated. Is smoothed by a Gaussian distribution function having a predetermined standard deviation, and the smoothed image is secondarily differentiated to calculate a secondary differential image. A contour pixel detection step of detecting a contour pixel from the obtained secondary differential image to obtain a contour pixel; a contour pixel obtained in the contour pixel detection step is set as a first value; A binarization step for replacing the binarized image as a binary value; the binarized image obtained in the binarization step is divided into blocks having a predetermined number of pixels, and a plurality of binarized image blocks are divided. Get Block dividing step, a counting step of counting the number of pixels having the first value for each binarized image block obtained in the block dividing step, and a block obtained in the counting step, A distorted block determination step of determining the binarized image block in which the number of pixels of the first value is equal to or less than a predetermined value as a distorted block, and smoothing each pixel in the distorted block obtained in the distorted block determination step. And a smoothing step for obtaining an image with reduced noise.