JP2003219419A - Image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enable an image processor to effectively reduce block noise with a smaller computation complexity at the time of constituting the processor to encode and decode images by blocks. <P>SOLUTION: This image processor is provided with a means of calculating the characteristic value of the blocks of an image at every block, a means which decides the correcting method of the blocks at every block, and a means which corrects pixel values in a block which is decided to be corrected by the decided correcting method by using the average pixel value of the block and blocks existing in the circumference of the block following the correcting method. This processor is also provided with a means which corrects the pixel values of the boundary pixels of the corrected block by using the pixel values of the boundary pixels and pixels existing in the circumference of the boundary pixels. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for removing block noise existing in an image coded / decoded in JPEG format or the like, and particularly, with a small calculation amount,
Moreover, the present invention relates to an image processing device that can effectively reduce block noise.

[0002]

2. Description of the Related Art In recent years, digital color images have been frequently found due to the spread of digital cameras. This image is stored in a coding / decoding format called "JPEG", for example. This is because using JPEG, the file size is 1 / 2-1 / compared with the uncompressed image.
This is because it can be compressed to about 20 and is suitable for storage and communication.

This encoding / decoding format is usually performed in the frequency domain. This will be specifically described with reference to FIG.

First, as shown in FIG. 30, the image is 8
It is divided into blocks of × 8 pixels, and each block is transformed into a frequency space by orthogonal transformation. As a result, a total of 64 coefficients relating to the high frequency component and the low frequency component are calculated.

Next, as shown in FIG. 30, quantization is performed by dividing the coefficient by a quantization threshold value (quantization table, common to the entire image) determined for each frequency. The quantization threshold value is set small in a low frequency region where humans are sensitive and set large in a high frequency region. This quantization process causes a part of information to be lost, which causes noise described later.

Finally, as shown in FIG. 30, data redundancy is removed by entropy coding to generate coded data. The decoder reproduces the coded image through inverse quantization and inverse transformation.

However, if the image data is compressed at a high compression rate during storage, various distortions (for example, block noise, mosquito noise, etc.) will occur in the image, and the image quality will be significantly deteriorated. Among them, block noise is particularly noticeable because it appears that a block-shaped contour appears in a flat image portion such as “sky”.

This is because the frequency component of the block becomes almost only the low frequency component due to the high compression, so that the block becomes a flat block when inversely quantized.

As a first conventional technique for making the block noise inconspicuous, as shown in FIG. 31, a pixel facing a block boundary (hereinafter abbreviated as a block boundary pixel or a boundary pixel) is smoothed. Technology that reduces the level difference at the block boundary by filtering to make block noise inconspicuous (Japanese Patent Laid-Open No. 2-5706
No. 7 “Image coding processing method”).

As a second conventional technique, smoothing filtering is performed in a wide range around a block boundary to reduce the difference between the level difference at the block boundary and the level difference at other than the block boundary. There is a technique for making the level difference at the boundary inconspicuous (Japanese Patent Laid-Open No. 10-22479).
No. 0 "Filter and filter method for removing blocky noise in compressed and expanded image".

In the second prior art, before filtering, the activity / average value of the area to be filtered is calculated, only the flat part is extracted based on these values, and this is filtered. Applicable range. The reason why only the flat portion is set as the filter application range is to prevent sharp lines and contours that are features in the image from being blurred by the filtering.

The activity is the non-uniformity of pixel values in the area. It is considered that the more non-uniform the image, the more abrupt the pixel value changes, that is, the contour / line exists in the region. Therefore, the contour / line is prevented from blurring by not performing filtering on a highly active area.

As shown in FIG. 32, the concrete processing procedure of the second conventional technique is to form flat first and second groups with a block boundary as a boundary, and to calculate an average value / activity for each group. Is calculated, based on these, it is determined whether the line / contour is blurred even if filtering is performed across both groups, and after setting the filter application range,
As shown in FIG. 33, a linear interpolation filter is applied to the entire set filter application range for smoothing.

[0014]

SUMMARY OF THE INVENTION In the first prior art,
Since filtering is performed only on the block boundary pixels, as shown in FIG. 34, there is a problem that only the level difference at the boundary may be conspicuous for the block noise manifested in the flat portion. is there.

It is considered that this is because there is no change in the level difference except at the block boundary, although the level difference at the block boundary is large. In the flat part, the part where the level difference does not change at all except the block boundary is distributed over a wide range, so that the level difference at the boundary part becomes more conspicuous.

On the other hand, in the second conventional technique, smoothing filtering is performed in a wide range around the block boundary in order to remove the block noise manifested in such a flat portion. As shown in, the level difference at the block boundary becomes inconspicuous.

However, the second conventional technique has a problem that a different correction amount must be calculated for every pixel in the filter application range.

Further, in the second conventional technique, since it is necessary to perform the noise removal processing on the block boundaries facing the vertical direction in the image and then perform the horizontal processing, the filter application range is detected. There is a problem in that it is necessary to perform the average value / activity calculation process for each column again in the horizontal direction.

Therefore, because of these problems, the second
According to the conventional technique of No. 1, there is a problem that the block noise removal process becomes slow.

The present invention has been made in view of the above circumstances, and has a configuration in which an image is block-encoded and decoded, with a small amount of calculation, and
It is an object of the present invention to provide a new image processing technique that can effectively reduce block noise.

[0021]

According to the present invention, when an image is block coded and decoded, a filter is applied to remove block noise existing in a flat image portion such as the sky. It is an object to solve the problem of the conventional technique that the calculation amount increases by expanding the range.

In order to solve this problem, the present invention encodes an image in order to reduce block noise generated when the original image is block-encoded / decoded.
A dividing unit that divides into the same block as that at the time of decoding or a different block, a calculating unit that calculates the characteristic value of each block of the image, and each block of the image based on the calculated characteristic value Each time, a decision is made to determine the correction method for that block (describe whether it is a block to be corrected and, if correction is to be performed, which peripheral block is used to perform the correction). Means and a first correction for correcting each pixel value in the block determined to be corrected by the correction method using the average pixel value of the block and blocks existing around the block according to the correction method Second means for correcting the pixel value of the boundary pixel of the corrected block using the pixel value of the boundary pixel and the pixel existing around the boundary pixel Configured to include a correction means.

Then, when adopting this configuration, the second correction means sets the pixel value of the correction target pixel as the correction target not only for the boundary pixel but also for the pixel located in the vicinity of the boundary pixel. In some cases, the pixel value of a pixel and a pixel existing around the pixel to be corrected may be used for correction.

When adopting this configuration, the calculating means may calculate, as the characteristic value of the block, a value indicating whether or not the block is flat and an average pixel value of the block. In response to the characteristic value, the determining means uses a correction method that is not a flat block and a flat block that does not have a flat block in which the difference in average pixel value is equal to or less than a specified threshold value. With respect to and, processing is performed so as to determine as a block that is not corrected, or to determine a flat block in which the difference between average pixel values existing in the periphery is equal to or less than a specified threshold as a peripheral block used when performing correction. Sometimes.

Here, the first correction means calculates an average value of average pixel values of the block which is determined to be corrected and the block to be corrected existing around the block, and each pixel in the block is calculated. By replacing the value with that, it may be processed so as to correct each pixel value in the block determined to be corrected.

Further, the first correction means is a difference value between the average pixel value of the block determined to be corrected and the average pixel value of the correction target block existing around the block, and the average pixel value of the block. May be calculated, and the difference value may be added to each pixel value in the block to process each pixel value in the block determined to be corrected.

Further, the second correcting means, when only the boundary pixel is the pixel to be corrected, the pixel value of the boundary pixel and the pixel value of the pixel next to the boundary pixel across the block boundary. , The processing may be performed so as to calculate the correction amount of the boundary pixel by referring to the pixel value of the adjacent pixel in the opposite direction.

When the boundary pixel and the pixels adjacent to the boundary pixel are the correction target pixels, the second correction means determines the pixel value of the correction target pixel and the direction in which the block boundary is seen from the correction target pixel. The pixel value of a specified number of pixels present in
The pixel value of the same number of pixels in the opposite direction may be referred to and the correction amount of the pixel value of the correction target pixel may be calculated.

Each processing means constituting the present invention described above can be realized by a computer program,
This computer program can be recorded in a recording medium such as a semiconductor memory and provided.

According to the present invention having such a configuration, (1)
As shown in FIG. 1, for each block, such as the characteristic value of the block (the average pixel value of the block, the standard deviation value of the pixel values of the block, the AC component obtained when the block is orthogonally transformed, etc.). (Activity) is calculated, and it is used to evaluate whether the block is flat and the difference in average pixel value between adjacent flat blocks, to determine the block correction method (whether to perform noise removal processing). Whether or not and which peripheral flat block is used to perform the noise removal processing) are determined.
(2) Subsequently, for each block to be corrected, an average value with the DC component of the peripheral block (a component indicating the height of the block, an average value of the entire block) is calculated for each block while following the instruction of the correction method. By correcting the DC component in such a manner, the step difference between the blocks is reduced as shown in FIG. 2, and (3) finally, the block boundary is corrected as shown in FIG. The block noise is processed by filtering with a fixed-size filter at the center.

In the present invention, according to the processing of (2) above,
The processing is performed so that the DC component is changed so that the step at the block boundary becomes small.

Since the step is already small at this point, the slope on the boundary becomes gentler than when only filtering is performed. According to the present invention, it is possible to reduce the difference between the level difference at the block boundary and the level difference near the block boundary by giving the gentle inclination in this way, and the boundary part becomes inconspicuous.

In the present invention, similarly to the second prior art, the pixel inside the block (hereinafter abbreviated as the pixel inside the block) is also corrected, but this correction is for the DC component. However, the correction amounts are all the same as shown in FIG. Therefore, according to the present invention,
There is no need to perform different calculations for each pixel in the block as in the prior art.

Therefore, according to the present invention, there is a case in which the image is block-encoded and decoded,
Block noise can be effectively reduced with a small amount of calculation.

In addition, as described above, in the second conventional technique, it is necessary to perform the noise removal processing on the block boundaries facing the vertical direction in the image and then perform the horizontal processing, so that the filter application is performed. It is necessary to perform the average value / activity calculation processing for detecting the range again for each column in the horizontal direction.

On the other hand, since the processing of (1) above is based on the characteristic value calculated for each block, according to the present invention, even if the processing direction changes, each block obtained previously is determined. Since it is possible to use the characteristic value of, it is not necessary to recalculate.

Also from this point, according to the present invention, the block noise can be reduced with a small amount of calculation.

[0038]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, the present invention will be described in detail according to embodiments.

FIG. 5 is a diagram for explaining removal of block noise according to the first embodiment of the present invention.

Here, it is assumed that an embodiment is used when a certain color image is output by a printer.

The printer 120 according to this example embodiment includes:
It has a block noise removal function, and automatically removes noise in the image before printing to the printer.

This example embodiment requires a color image 100, a PC 110 and a printer 120.

The color image image 100 is composed of header data 101 and pixel data 102, and is stored in the PC 110. As shown in FIG. 6, the color image 100 is coded / decoded in 8 × 8 block units, and is composed of n blocks.

The PC 110 has a pixel data transfer function 111.
And transfers only the pixel data 102 to the printer 120 at the time of printer output.

The printer 120 has a noise removing function 1 for the pixel data 102 transferred from the PC 110.
Block noise in the image is removed by 21 and the image is printed out.

The noise removal function 121 has an average value / standard deviation value calculation module 122, a correction method determination module 123 and a correction processing module 124 in order to remove block noise in the image.

FIG. 7 shows the overall flow of the processing procedure of the first embodiment, FIG. 8 shows the detailed flow of the processing of the average value / standard deviation value calculation module 122, and FIG.
FIG. 10 illustrates a detailed flow of processing of the correction method determination module 123, and FIG. 10 and FIG.
The detailed flow of the process of 24 is illustrated.

Next, the block noise reduction process realized in the first embodiment will be described in detail in the order of the flow of FIG.

(1) Process 200 The PC 110 transfers the pixel data 102 to the printer 120.

(2) Step 201 (FIG. 8 shows its detailed processing) When the printer 120 receives the pixel data 102 from the PC 110, the pixel data 102 is divided into blocks as shown in the flowchart of FIG. , Each block i (i =
1 to n), the average value Ave _i and the standard deviation value Stv _i of the pixel data 102 of each block _i are calculated.

(3) Step 202 (FIG. 9 shows its detailed processing) When the printer 120 calculates the average value Ave _i and the standard deviation value Stv _i of each block i, each block i is processed as shown in the flowchart of FIG. The standard deviation value Stv _{i of i} is compared with a threshold value TH1 (for example 5.0). If the standard deviation value Stv _i is greater than or equal to the threshold value TH1, it means that the activity is high, that is, the contour or line exists as described above. Therefore, in order not to blur this, the value “0” indicating that noise removal is not performed for this block i is substituted for the correction method determination value Flag _i .

On the other hand, if the standard deviation value Stv _i is less than the threshold value TH1, it is determined whether or not noise can be removed based on the left and right blocks. By the way, when the block i is at the right end in the image, it is determined whether the noise removal can be performed based on only the left block, and when it is at the left end, the noise removal is performed based on only the right block. If it is not at the right end or the left end, it is determined whether noise removal can be performed based on the left and right blocks. Here, the case where the block i is neither the left end nor the right end will be described as an example.

[0053] The absolute value of the difference between the average value Ave _i of the average value Ave _i-1 and the block i of the left block i-1 is TH2 (e.g.
13.0) or more, the left block i-1 is regarded as a block forming a different object and is not used as a reference for correction processing. Similarly, the standard deviation Stv of the left block i-1
_{Even when i-1} is equal to or greater than TH1, the left block i-1 is regarded as a block forming a different object and is not used as a reference for correction processing.

The average value Ave of the right block i + 1
The absolute value of the difference between _{i + 1} and the average value Ave _{i of} block i is TH2.
If it is (for example, 13.0) or more, the right block i + 1
Are regarded as blocks forming different objects and are not used as a reference for correction processing. Similarly, when the standard deviation Stv _{i + 1} of the right block i + 1 is equal to or more than TH1, the right block i + 1
+1 is regarded as a block forming a different object and is not used as a reference for correction processing.

In this way, the printer 120 is operated as shown in FIG.
By performing the determination process according to the flowchart of FIG.
For the correction method decision value Flag _{i in the} form as shown in FIG.
Assign a value to.

(3) Step 203 (Detailed Processing in FIGS. 10 and 11) The printer 120 performs a correction process for removing block noise based on the value of Flag _i calculated in step 202.

First of all, according to the flow chart of FIG.
Correction amount for the entire lock Correct _i(Complement of block i
Positive amount) is calculated.

That is, (a) When the value of Flag _i is 0, the correction amount Correct _i is not calculated.

(B) When the value of Flag _i is 1, the correction amount Correct _i is expressed as Correct _i = (Ave _i-1 + Ave _i + Ave _{i + 1} ) / 3-Av
It is calculated according to the formula e _i .

(C) When the value of Flag _i is 2 The correction amount Correct _i is calculated according to the following formula: Correct _i = (Ave _i-1 + Ave _i ) / 2-Ave _i .

(D) When the value of Flag _i is 3 The correction amount Correct _i is calculated according to the following formula: Correct _i = (Ave _i + Ave _{i + 1} ) / 2-Ave _i . Then, the correction amount Correct _i (correction amount of block i) for the entire block is calculated.

Then, the calculated correction amount Correct
_i is added to the pixel values of all pixels in block i.

The meaning of this correction is to uniformly increase or decrease the pixel value of each pixel of the block i so as to approach the average pixel value of the peripheral block, thereby making it possible to obtain the same value as that of the peripheral block. It is processed so as to reduce the step.

After the printer 120 completes the correction for all the blocks, the printer 120 of FIG.
The block boundary pixels facing the boundaries of each block are filtered according to the processing of the flowchart of FIG.

That is, (a) When both Flag _i-1 and Flag _i are not 0, the block boundary pixel facing the boundary between block i-1 and block i is filtered.

(B) When both Flag _i and Flag _{i + 1} are not 0, the block boundary pixel facing the boundary between block i and block i + 1 is filtered. In this way, the block boundary pixels facing the boundary of each block are filtered.

As shown in FIG. 13, pixels A _j-1, A _j, A _{j + 1,}
A _{j + 2} is defined and the pixel values of these pixels are L
_{If j-1,} L _j, L _{j + 1, and} L _{j + 2} are used, the filtering method performed at this time is, for example, (a) the corrected block boundary pixel for the block boundary pixel A _j of the block i-1. The pixel value L _j of A _j is corrected according to the formula: L _j ^' = (L _j-1 + L _j + L _{j + 1} ) / 3.

(B) The pixel value L _{j + 1} of the corrected block boundary pixel A _{j + 1 for} the block boundary pixel A _{j + 1} possessed by the block i is L _{j + 1} ^' = (L _j + L _{j + 1} + L Correct according to the formula _{j + 2} ) / 3. In this way, the block boundary pixels facing the boundary of each block are filtered.

As will be described later, this correction makes it possible to give the block boundary a gentle slope.

The processes of steps 202 and 203 described above are similarly performed in the vertical direction.

(4) Step 204 After the correction process is completed, the actual printing is performed and the process is terminated.

In this way, the printer 120 according to the present embodiment operates so as to automatically remove the block noise in the image before the printer output.

In this embodiment, filtering is performed only on block boundary pixels.

However, as a matter of course, in order to give a more gradual inclination, not only the block boundary pixel but also one or a plurality of blocks located inside the block boundary pixel to the extent that the processing speed is not extremely decreased. It is also effective to perform filtering on inner pixels.

Next, another embodiment of the present invention will be described.

FIG. 14 is a diagram for explaining removal of block noise according to the second embodiment of the present invention.

Here, it is assumed that the embodiment is an example in which a certain color image is output by a printer.

This example embodiment requires color image 300 and PC 310.

The color image image 300 is composed of header data 301 and pixel data 302, and is stored in the PC 310. As shown in FIG. 15, the color image 300 is coded / decoded in 8 × 8 block units, and is composed of a total of n × m blocks of n rows and m columns.

The PC 310 has a noise removing function 311. The noise removing function 311 removes block noise in an image and displays it on the display.

The noise removing function 311 has an average value / activity calculation module 312 and a correction processing module 313 in order to remove block noise in the image.

FIG. 16 shows the overall flow of the processing procedure of the second embodiment, and FIGS. 17 and 18 show the average value /
A detailed flow of the processing of the activity calculation module 312 is illustrated, and FIGS. 19 and 20 illustrate a detailed flow of the processing (the correction method is also determined) of the correction processing module 313.

Next, in the order of the flow of FIG.
The block noise reduction process realized in the embodiment will be described in detail.

(1) Step 400 (Detailed Processing in FIGS. 17 and 18) The PC 310 firstly, as shown in the flowchart of FIG. 17, the AC component of each block (j, k) from the header data 301. Then, for each block (j, k), those AC components are compared with the threshold value TH1 (for example, 5.0). If any AC component exceeds the threshold value, it is a block with high activity. In consideration of this, the value “0” indicating that processing is not performed for this block (j, k) is set to the correction method determination value Flag
Substitute for _{(j, k)} (initial value is 1).

Then, the PC 310 divides the pixel data 302 into blocks as shown in the flowchart of FIG. 18, and for each block (j, k), the pixel data held by each block (j, k) is divided. The average value Ave _{(j, k)} of 302 is calculated.

(2) Step 401 (Detailed Processing in FIGS. 19 and 20) The printer 120 subsequently performs the correction processing while determining the correction method according to the flowcharts in FIGS. 19 and 20.

The printer 120 first executes the processing shown in the flowchart of FIG.

That is, the correction process is not performed for the block (j, k) whose Flag _{(j, k)} value is 0.

If the value of Flag _{(j, k)} is 1, it is determined whether the upper, lower, left and right blocks serve as a reference for the correction process.
Here, the block (j, k) that is not located at the upper end, the lower end, the left end, or the right end in the image will be described as an example.

(A) When Flag _{(j-1, k)} is 0,
The upper blocks (j-1, k) are not used as reference for the correction process. Similarly, the absolute value of the difference between the average value of the average value of the upper block (j-1, k) Ave (j-1, k) and block _{(j, k) Ave (j} , k) is TH
If it is 2 (for example, 13.0) or more, the upper block (j
-1, k) are regarded as blocks forming different objects and are not used as a reference for correction processing.

When the correction process is not used as a reference, 1 is subtracted from Brock-Num (the initial value is 5) indicating the number of blocks to be averaged, and when it is used as a reference, the total sum of the average values is managed.
To Ave-Sum (initial value is Ave _{(j, k) )} , add average value Ave _{(j-1, k)} which is a reference for correction.

(B) When Flag _{(j + 1, k)} is 0,
The lower block (j + 1, k) is not used as a reference for correction processing. Similarly, the absolute value of the difference between the mean values of the lower block (j + 1, k) Ave and _{(j + 1, k)} block (j, k) mean value Ave _{(j, k)} of the TH
When it is 2 or more, the lower block (j + 1, k) is regarded as a block forming a different object and is not used as a reference for the correction process.

When the correction process is not used as a reference, 1 is subtracted from Brock-Num which indicates the number of blocks to be averaged. When it is used as a reference, Ave-Sum for managing the sum of average values is used. , Add the average value Ave _{(j + 1, k)} as a reference for correction.

(C) When Flag _{(j, k-1)} is 0,
The block (j, k-1) on the left is not used as a reference for correction processing. Similarly, the absolute value of the difference between the average value Ave _{(j, k-1)} of the left block (j, k-1) and block (j, k) mean value Ave _{(j, k)} of the TH
If it is 2 or more, the left block (j, k-1) is regarded as a block forming a different object and is not used as a reference for the correction process.

When the correction process is not used as a reference, 1 is subtracted from Brock-Num which indicates the number of blocks to be averaged, and when it is used as a reference, Ave-Sum for managing the sum of average values is used. , Add the average value Ave _{(j, k-1)} as a reference for correction.

(D) When Flag _{(j, k + 1)} is 0,
The block (j, k + 1) on the right is not used as a reference for correction processing. Similarly, the absolute value of the difference between the right block (j, k + 1) Mean value Ave _{(j, k + 1)} between the block (j, k) mean value Ave _{(j, k)} of the TH
When it is 2 or more, the right block (j, k + 1) is regarded as a block forming a different object and is not used as a reference for the correction process.

When the correction process is not used as a reference, 1 is subtracted from Brock-Num which indicates the number of blocks to be averaged, and when it is used as a reference, Ave-Sum for managing the sum of average values is used. , Add the average value Ave _{(j, k + 1)} as the reference for correction.

Thus, when Brock-Num and Ave-Sum are obtained for each block (j, k), the printer 1
20 subsequently performs a correction process for removing block noise according to the process of the flowchart of FIG. However, Fl
The correction process described below is not performed for the block (j, k) in which the value of ag _{(j, k)} is 0.

Although not shown in the flow chart of FIG. 20, the printer 120 first corrects the entire block according to the calculation formula Correct _{(j, k)} = (Ave-Sum) / (Brock-Num). Quantity C
Calculate orrect _{(j, k)} (correction amount of block (j, k)) and replace the calculated correction amount Correct _{(j, k)} as the correction result of all pixels in block (j, k) .

After the printer 120 completes the correction for the entire block for all blocks, the printer 120 of FIG.
The block boundary pixels facing the boundaries of each block are filtered according to the processing of the flowchart of FIG.

That is, (a) When both Flag _{(j-1, k)} and Flag _{(j, k)} are not 0, a value near the boundary between block (j-1, k) and block (j, k) is Filter the pixels.

(B) When both Flag _{(j + 1, k)} and Flag _{(j, k)} are not 0, to the pixel near the boundary between block (j + 1, k) and block (j, k) Apply filtering processing.

(C) If Flag _{(j, k-1)} and Flag _{(j, k)} are not 0, go to the pixel near the boundary between block (j, k-1) and block (j, k). Apply filtering processing.

(D) When both Flag _{(j, k + 1)} and Flag _{(j, k)} are not 0, to the pixel near the boundary between block (j, k + 1) and block (j, k) Apply filtering processing. In this way, the pixels near the boundary of each block (j, k) (block boundary pixels and, for example, one pixel from the block boundary pixel to the center of the block) are filtered.

As shown in FIG. 21, pixels A _j-3, A _j-2, A
_j-1, A _j, A _{j + 1,} A _{j + 2,} A _{j + 3,} A _{j + 4} are defined, and the pixel values of these pixels are L _j-3, L _j-2, L _{j, respectively. -1,} L _j, L _{j + 1,}
If L _{j + 2,} L _{j + 3 and} L _{j + 4} are used, the filtering method in the horizontal direction performed at this time is, for example, (a) the pixels A _j-1, A of the block (j, k-1) the corrected pixel value L _j-1, L _j of _{j, L j-1 '=} (L j-3 + L j-2 + L j-1 + L j + L j + 1) /
5 L _j '= (L _j-2 + L _j-1 + L _j + L _{j + 1} + L _{j + 2} ) /
Correct according to formula 5.

(B) Pixel A _{j + 1,} A possessed by block (j, k)
_{j + 2,} the corrected pixel value _{L j + 1, L j +} 2 _{for, L j + 1 '= (} L j-1 + L j + L j + 1 + L j + 2 + L j + 3) /
5 L _{j + 2} '= (L _j + L _{j + 1} + L _{j + 2} + L _{j + 3} + L _{j + 4} ) /
Correct according to formula 5. In this way, the horizontal filtering is performed.

As will be described later, this correction makes it possible to give the block boundary a gentle slope.

Then, as shown in FIG._k-3,
A_k-2,A_k-1,A_k,A_{k + 1,}A_{k + 2,}A_{k + 3,}A_{k + 4}Define
The pixel value of these pixels is L_k-3,L_k-2,L_k-1,L
_k,L _{k + 1,}L_{k + 2,}L_{k + 3,}L_{k + 4}If you say
The vertical filtering method is, for example, (B) Pixel A held by block (j-1, k)_k-1,A_kabout
Correction of Pixel value L_k-1,L_kTo L_k-1'= (L_k-3+ L_k-2+ L_k-1+ L_k+ L_{k + 1}) /
5 L_k'= (L_k-2+ L_k-1+ L_k+ L_{k + 1}+ L_{k + 2}) /
5 It corrects according to the formula.

(B) Pixels A _{k + 1,} A possessed by block (j, k)
The corrected pixel value _{L k + 1, L k +} 2 for _{k + 2, L k + 1} '= (L k-1 + L k + L k + 1 + L k + 2 + L k + 3) /
5 L _{k + 2} '= (L _k + L _{k + 1} + L _{k + 2} + L _{k + 3} + L _{k + 4} ) /
Correct according to formula 5. The vertical filtering is performed as follows.

As will be described later, this correction makes it possible to give the block boundary a gentle slope.

In this way, the PC according to the present embodiment example is
The 310 operates so as to automatically remove block noise in the image and then display it on the display.

Next, it will be verified that by using the present invention, it becomes possible to remove block noise while reducing the amount of calculation as compared with the prior art.

In the present invention, first, the characteristic value is calculated for each block, and it is judged whether the block noise is actualized or not based on this characteristic value. For blocks in which block noise is manifested, DC component correction of blocks and fixed size filtering of block boundaries are performed.
All pixel values in the image are used once for this characteristic value calculation.

On the other hand, for example, in FIG.
When the noise removal processing is performed on the block boundaries A and B using the conventional technique of No. 1, the characteristic values (activity and average value) of the four groups C, D, E, and F facing the block boundaries are calculated. There is a need. At that time, the pixel G is used for calculation twice. In other words, all pixels in the image are used once in each direction, twice in total, for the purpose of calculating the activity in the horizontal and vertical directions.

For example, as shown in FIG. 24, for an image in which 8 × 8 pixel blocks are arranged in the vertical direction and the horizontal direction are n,
As in the present invention, when the average value / variance value is obtained for each block, as in the second conventional technique (see FIG. 30),
The calculation amount for obtaining the average value / variance value for the first and second groups that cross x block boundaries in each row / column is shown below.

Incidentally, the number of pixels in each group in the second conventional technique is 8n / 2x because there are x boundaries in one side 8n (FIG. 24). The total number of groups is
Since there are 2x groups in one row and the number of rows is 8n, the number is 2x × 8n, and since there are the same number in the vertical direction, the total is 32nx (2x × 8n × 2).

The average value (Block-Ave) and the variance value (Block-Var / Block-Stdev) of the block are

[0118]

[Equation 1]

It is calculated by Here, a _i is each pixel value.

The calculation amount of the average value (Block-Ave) is 63 additions in the numerator and 1 division thereafter. On the other hand, the variance value
The calculation amount of (Block-Var / Block-Stdev) is 2 times subtraction and 1 multiplication because the average value is subtracted from each pixel in the numerator part and squared.
The total number of times is 64 times, and finally 63 times are added. Then, the denominator part is subtracted once, and then the denominator is divided by the numerator once.

Average value (Group-Ave) and variance value (Group-Var / Block-Stdev) of one group in the second conventional technique.
Is

[0122]

[Equation 2]

It is calculated by Here, a _i is each pixel value.

The calculation amount of the average value (Group-Ave) is one addition (8n / 2x) -1 in the numerator, and one division thereafter. On the other hand, the calculation amount of the variance value (Group-Var / Block-Stdev) is subtracted because the average value is subtracted from each pixel and squared in the numerator part.
Perform one multiplication and one multiplication (8n / 2x) times, and finally add (8n / 2x) -1 times to them. The denominator part is subtracted once, and then the denominator is divided by the numerator once. These are done twice in the horizontal and vertical directions.

When it is performed for each block, as can be seen from FIG. 25, according to the present invention, the number of boundaries x is the maximum n−1, so that the calculation amount is surely reduced (to about half the maximum). Decrease).

Next, in the present invention, the level difference at the block boundary is reduced in advance by performing a constant correction on the entire block.

Here, the correction amount Block-Ave ' _{(j, k)} for the entire block (j, k) is Block-Ave' _{(j, k)} = 4 × Block-Ave _{(j, k)} / 5- Block
-Ave _{(j-1, k)} / 5-Block-Ave _{(j, k-1)} / 5-Block-A
ve _{(j, k + 1)} / 5- Block-Ave _{(j + 1, k)} / 5 However, Block-Ave _{(j-1, k)} : Average value of block (j-1, k) Block-Ave _{(j, k-1)} : Average value of block (j, k-1) Block-Ave _{(j, k)} : Average value of block (j, k) Block-Ave _{(j, k + 1)} : Block (A Average value of j, k + 1) Block-Ave _{(j + 1, k)} : It is the average value of block (j + 1, k). However, the positional relationship of the blocks is assumed to be that shown in FIG.

As can be seen from this equation, the calculation of the correction amount for the entire block is completed by one multiplication, four subtractions and one division.

Next, in the present invention, the correction amount for the entire block is subtracted from the pixel value in each block.

Next, in the present invention, in order to make the difference between the level difference at the block boundary and the level difference other than the block boundary inconspicuous, filtering is performed on the block boundary pixel while referring to each pixel value in the vicinity. , Give a gentle slope to the block boundary.

Here, assuming that filtering for each pixel is equivalent to the conventional one, and the calculation amount of filtering is both A, the calculation amounts for the in-block pixel and the block boundary pixel are as shown in FIG.

As described above, the calculation amount for the pixels in the block is obtained by simply subtracting the correction amount for the entire block from the pixel value in each block. Therefore, according to the present invention, the calculation amount is greatly reduced as compared with the conventional technique. It will be possible.

Further, the present invention will be described.

(1) In the present invention, the noise removal processing is performed only on the block noise by using the difference from the average pixel value of the surrounding blocks.

When noise is removed, if the DC components are averaged between blocks forming different objects, the colors of both objects are mixed. Therefore, in the present invention, an average value for each block is obtained to prevent this. Is adopted and compared with the average value of surrounding blocks.

Then, as a result of the comparison, when the difference from the average value of the surrounding blocks is extremely large, it indicates that the colors of the blocks are greatly different, which indicates that the surrounding blocks form different objects. Is likely to
The surrounding blocks are excluded from the block noise removal processing.

(2) In the present invention, the noise removal processing is performed only on the block noise by using the activity of the block.

It is highly possible that a block having high activity forms a contour / line that is a feature in the image. If you filter the contours and lines, they will appear blurred. Therefore, in the present invention, the activity is calculated for each block, and when there is a block with high activity,
Noise removal processing is not performed on such blocks.

(3) In the present invention, in addition to calculating the variance and standard deviation as the activity of the block, if the image data is coded / decoded in the frequency domain, the AC component at the time of decoding is calculated. The activity is calculated based on this. When such activity is used, the calculation amount can be significantly reduced.

For example, in the encoding / decoding process,
It is assumed that the coefficient after inverse quantization has a value as shown in FIG. Only the upper left value (coefficient) is a DC component, and the other 63 values (coefficients) are AC components. The DC component indicates the average value of pixel values of each block. The AC component indicates something like the amplitude of each frequency from low frequency to high frequency.

A large frequency amplitude means that the block forms a sharp line / contour. Therefore, if any of the AC components is equal to or greater than the threshold TH1, the block is not filtered.

FIG. 28 shows a state in which the AC component and the threshold value TH1 are sequentially compared with each other like. Here, since the coefficient is TH1 or more as a result of the comparison, the processing is ended here, and it is decided not to perform noise removal for this block.

As described above, in the present invention, the activity is calculated (measured) by using each AC component of the block already calculated at the time of encoding without calculating the variance / standard deviation of all pixels in the block. It is also possible to process as follows.

(4) In the present invention, for the pixels inside the block to be corrected, the correction amount is calculated from the average pixel value of the block and blocks existing around the block. .

In the present invention, for the pixels inside the block to be corrected, the amount of calculation is reduced by performing DC correction instead of individual filtering.

The calculation amount can be reduced by reducing the number of times of individual filtering applied to the boundary pixel and its neighboring pixels as much as possible. Further, in order to achieve both of the provision of the gentle inclination to the boundary portion, only the portion of the pixels in the block close to the boundary of the block may be filtered.

(5) In the present invention, processing is performed so that a smooth boundary is given to the block boundary portion.

In order to give a smooth slope to the block boundary, it is necessary to correct the step difference between blocks due to block noise so that it is divided into several pixels near the boundary.

At this time, the closer the pixel is to the block boundary, the larger the correction amount is made to match the level value of the adjacent block, and the farther the boundary is, the smaller the correction amount is. Realize the provision of. In order to realize a relatively small amount of calculation, as described above, reference pixels having the same number on the left and right sides are used.

That is, as shown in FIG.
_j-3,A_j-2,A_j-1,A_j,A_{j + 1,}A_{j + 2,}A _{j + 3,}A_{j + 4}Define
The pixel values of these pixels to L_j-3,L_j-2,L
_j-1,L_j,L_{j + 1,}L_{j + 2,}L_{j + 3,}L_{j + 4}If so, amend
The target pixel is A_j-1,A_j,A_{j + 1,}A_{j + 2}And against these
Correction result L_j-1', L_j', L_{j + 1}', L_{j + 2}'           L_j-1'= (L_j-3+ L_j-2+ L_j-1+ L_j+ L_{j + 1}) / 5           L_j'= (L_j-2+ L_j-1+ L_j+ L_{j + 1}+ L_{j + 2}) / 5           L_{j + 1}'= (L_j-1+ L_j+ L_{j + 1}+ L_{j + 2}+ L_{j + 3}) / 5           L_{j + 2}'= (L_j+ L_{j + 1}+ L_{j + 2}+ L_{j + 3}+ L_{j + 4}) / 5 It is calculated as follows.

The correction target pixel A _j-1 is used when the correction amount is determined using the five pixel values, and there are four pixels (A _j-3, A _j-2, A in the block 1). _{Since the} correction amount is calculated based on ( _j-1, A _j ), the correction result is close to the average value of block 1. On the other hand, the correction target pixel A _j is in the block 1 when the correction amount is determined using the five pixel values.
Since the calculation is performed on the basis of the three pixels (A _j-2, A _j-1, A _j ) in the block, the correction result for A _j-1 is closer to the average value of block 2.

As described above, if the number of pixels to be referred to is the same on the left and right, the shorter the distance between the pixel to be corrected and the block boundary, the larger the correction amount. Therefore, in the present invention, a simple formula is used to calculate the block 1 It is possible to provide a gradual slope that moves from the average value to the average value of block 2.

(6) In the present invention, the calculation amount is as small as possible,
Processing is performed so as to give a smooth slope to the block boundary.

In order to suppress the amount of calculation for filtering as much as possible, it is necessary to reduce the number of pixels to be referred to as much as possible. On the other hand, in order to be able to give a smooth slope from the inside of a block to the inside of the other block, the pixel has a total of 3 pixels including the pixel in the block, the block boundary pixel, and the pixel of the block existing across the block boundary. It is necessary to refer to the type when filtering.

Therefore, in the present invention, the block boundary pixel, the pixel (pixel A) next to the block boundary pixel adjacent to the block boundary pixel, and the pixel A as seen from the block boundary pixel are opposite to each other. It is possible to suppress the calculation amount as much as possible by performing processing so that the average value of a total of 3 pixels in 1 is used as the correction result.

(7) In the present invention, processing is performed so that a smooth slope can be imparted to the block boundary portion in a short time.

According to the present invention, the noise removal processing in the horizontal / vertical directions can be performed at once by simultaneously referring to the upper, lower, left and right blocks.

(8) In the present invention, processing is performed so that a smooth slope can be given to the block boundary portion in a short time.

In the present invention, the pixel value in a block is replaced with a value calculated from the average pixel value of the block and blocks existing around the block.

If the activity of a block is very low, the result should not make a difference even if the calculated correction amount for the entire block is not subtracted for each block and is simply replaced. In this case, since it is not necessary to perform the subtraction process on each pixel, the calculation amount can be reduced.

(Supplementary Note 1) An image processing apparatus for reducing block noise generated when an original image is block-encoded / decoded, means for calculating a characteristic value of each block of an image, Based on the characteristic value, means for determining a correction method for each block of the image, and pixel values in the block determined to be corrected by the correction method are set to the block according to the correction method. And a means for correcting using the average pixel value of blocks existing around the block, and a pixel value of the boundary pixel of the corrected block, the pixel value of the boundary pixel and the pixel value of the pixel existing around the boundary pixel. An image processing apparatus comprising:

(Supplementary Note 2) In the image processing apparatus according to Supplementary Note 1, the means for correcting the pixel value of the boundary pixel includes, in addition to the boundary pixel, a pixel located in the vicinity of the boundary pixel as a correction target. An image processing apparatus, which corrects a pixel value of a correction target pixel using the pixel values of the correction target pixel and pixels existing around the correction target pixel.

(Supplementary Note 3) In the image processing apparatus according to Supplementary Note 1 or 2, whether or not the determining means is a block for performing correction based on the characteristic value,
An image processing apparatus characterized by determining a correction method that describes which peripheral block is used for correction when performing correction.

(Supplementary Note 4) In the image processing apparatus according to Supplementary Note 3, the determining means is based on the characteristic value and is not a flat block, but is a flat block. An image processing apparatus characterized by determining that there is no flat block whose difference is less than or equal to a prescribed threshold value as a block that is not corrected.

(Supplementary Note 5) In the image processing apparatus according to Supplementary Note 3, the determining means determines a flat block in which a difference between average pixel values existing in the periphery is equal to or less than a specified threshold value based on the characteristic value. An image processing device characterized in that it is determined as a peripheral block to be used for correction.

(Supplementary Note 6) In the image processing apparatus according to any one of Supplementary Notes 1 to 5, the means for correcting each pixel value in the block calculates the average value of the average pixel values and corrects the average pixel value. An image processing apparatus characterized by correcting each pixel value in a correction target block by replacing each pixel value in the target block with it.

(Supplementary Note 7) In the image processing apparatus according to any one of Supplementary Notes 1 to 5, the means for correcting each pixel value in the block is the average value of the average pixel value and the average of the correction target block. Compensating each pixel value in the correction target block by calculating a difference value with the pixel value and adding the difference value to each pixel value in the correction target block,
A characteristic image processing device.

(Supplementary Note 8) In the image processing apparatus according to Supplementary Note 1, the means for correcting the pixel value of the boundary pixel includes the pixel value of the boundary pixel and the pixel next to the pixel adjacent to the boundary pixel across the block boundary. An image processing device characterized by calculating a correction amount of a pixel value of a boundary pixel by referring to a pixel value and a pixel value of an adjacent pixel in a direction opposite to the pixel value.

(Supplementary Note 9) In the image processing apparatus according to Supplementary Note 2, the means for correcting the pixel value of the correction target pixel exists in the direction in which the pixel value of the correction target pixel and the block boundary as viewed from the correction target pixel exist. An image processing apparatus characterized in that the correction amount of the pixel value of the correction target pixel is calculated with reference to the pixel values of the specified number of pixels and the pixel values of the same number of pixels in the opposite direction. .

(Supplementary Note 10) An image processing method for reducing block noise generated when block coding / decoding of an original image is performed, wherein a process of calculating characteristic values of each block of the image, For each block of the image based on the characteristic value, the process of determining the correction method of the block and each pixel value in the block determined to be corrected by the correction method are And a process of correcting using an average pixel value of blocks existing around the block, and a pixel value of a boundary pixel of the corrected block is a pixel value of the boundary pixel and a pixel existing around the boundary pixel. An image processing method, comprising the step of:

(Supplementary note 11) An image processing program for reducing block noise generated when the original image is block-coded / decoded, and a process of calculating the characteristic value of each block of the image. Then, based on the characteristic value, the process of determining the correction method for each block of the image, and the pixel value in the block determined to be corrected by the correction method are set according to the correction method. , A process of correcting using the average pixel value of the block and blocks existing around the block, and a pixel value of the boundary pixel of the corrected block, a pixel existing around the boundary pixel and the boundary pixel An image processing program for causing a computer to perform a process of correcting using the pixel value of.

(Supplementary Note 12) A recording medium of an image processing program, in which a program for reducing block noise generated when block encoding / decoding of an original image is recorded, is recorded for each block of an image. Processing for calculating the characteristic value of the pixel, processing for determining the block correction method for each block of the image based on the characteristic value, and each pixel in the block determined to be corrected by the correction method The value is corrected according to the correction method by using the average pixel value of the block and blocks existing around the block, and the pixel value of the boundary pixel of the corrected block is set to the boundary pixel and the boundary. A recording medium for an image processing program in which a program for causing a computer to perform a correction process using a pixel value of a pixel existing around a pixel is recorded. .

[0173]

As described above, according to the present invention, when the block noise manifested in the flat portion in the boundary image of the block boundary is removed, the correction and the filtering for the entire block are combined, so that the block boundary can be processed in a shorter processing time than before. A gentle slope can be imparted.

Thus, according to the present invention, there is a case where the image is block-coded and decoded,
Block noise can be effectively reduced with a small amount of calculation.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of the present invention.

FIG. 2 is an explanatory diagram of the present invention.

FIG. 3 is an explanatory diagram of the present invention.

FIG. 4 is an explanatory diagram of the present invention.

FIG. 5 is a first embodiment example of the present invention.

FIG. 6 is an explanatory diagram of blocks.

FIG. 7 is a processing procedure of the first embodiment.

FIG. 8 is a processing procedure of the first embodiment.

FIG. 9 is a processing procedure of the first embodiment example.

FIG. 10 is a processing procedure of the first embodiment example.

FIG. 11 is a processing procedure of the first embodiment example.

FIG. 12 is an explanatory diagram of a correction method determination value Flag.

FIG. 13 is an explanatory diagram of blocks.

FIG. 14 is a second embodiment example of the present invention.

FIG. 15 is an explanatory diagram of blocks.

FIG. 16 is a processing procedure of the second embodiment example.

FIG. 17 is a processing procedure of the second embodiment.

FIG. 18 is a processing procedure of the second embodiment.

FIG. 19 is a processing procedure of the second embodiment example.

FIG. 20 is a processing procedure of the second embodiment.

FIG. 21 is an explanatory diagram of blocks.

FIG. 22 is an explanatory diagram of blocks.

FIG. 23 is an explanatory diagram of blocks.

FIG. 24 is an explanatory diagram of a block.

FIG. 25 is a diagram for explaining a comparison of the calculation amount between the present invention and the prior art.

FIG. 26 is an explanatory diagram of a block.

FIG. 27 is a diagram for explaining a comparison of the calculation amount between the present invention and the prior art.

FIG. 28 is an explanatory diagram of the present invention.

FIG. 29 is an explanatory diagram of a block.

FIG. 30 is an explanatory diagram of an image processing technique using a block coding / decoding configuration.

FIG. 31 is an explanatory diagram of a conventional technique.

FIG. 32 is an explanatory diagram of a conventional technique.

FIG. 33 is an explanatory diagram of a conventional technique.

FIG. 34 is an explanatory diagram of a conventional technique.

FIG. 35 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

100 color image image 101 header data 102 pixel data 110 PC 111 pixel data transfer function 120 printers 121 Noise removal function 122 Average value / standard deviation value calculation module 123 Correction method determination module 124 Correction processing module 300 color image image 301 header data 302 pixel data 310 PC 311 Noise removal function 312 Average value / activity calculation module 313 Correction processing module

Continued front page    F term (reference) 5C059 KK03 MA00 MA23 MC11 MC31                       ME01 PP01 PP14 SS15 SS28                       TA69 TA70 TB08 TC02 TC10                       TC33 TC42 TD03 TD04 TD12                       UA05                 5C077 LL02 LL05 LL18 LL19 MP08                       NN04 PP09 PP31 PP46 PQ18                       TT02                 5C078 AA09 BA42 CA21 DA01 DA02

Claims (5)

[Claims] 1. An image processing apparatus for reducing block noise generated when an original image is block-encoded / decoded, comprising means for calculating a characteristic value of each block of an image, and the characteristic value. Based on the above, for each block of the image, a means for determining the correction method of the block, and the pixel values in the block determined to be corrected by the above correction method A means for correcting using an average pixel value of blocks existing around the block, and a pixel value of a boundary pixel of the corrected block using pixel values of the boundary pixel and pixels existing around the boundary pixel. An image processing apparatus characterized by comprising: 2. The image processing apparatus according to claim 1, wherein the means for correcting the pixel value of the boundary pixel corrects not only the boundary pixel but also the pixel located in the vicinity of the boundary pixel as the correction target. An image processing apparatus, characterized in that a pixel value of a target pixel is corrected by using the correction target pixel and a pixel value of a pixel existing around the correction target pixel. 3. The image processing apparatus according to claim 1, wherein the means for determining is a block for performing correction based on the characteristic value, and in the case of performing correction, An image processing apparatus characterized by determining a correction method that describes which peripheral block is used to perform correction. 4. The image processing apparatus according to claim 3, wherein the means for determining is not a flat block and a flat block based on the characteristic value.
An image processing apparatus characterized in that if there is no flat block having a difference in average pixel value equal to or less than a prescribed threshold value in the vicinity, it is determined as a block not to be corrected. 5. The image processing apparatus according to claim 3, wherein the determining unit determines a flat block in which a difference between average pixel values existing in the periphery is equal to or less than a specified threshold value based on the characteristic value. An image processing device characterized in that it is determined as a peripheral block to be used for correction.