JP4239893B2 - Block noise judgment method and program - Google Patents

Description

  The present invention relates to a block noise determination method and program, and more specifically, determines the degree of block noise that may occur at block boundaries in compression in a compressed image obtained by dividing and compressing an image into vertical and horizontal blocks having a predetermined number of pixels. The present invention relates to a block noise determination method and a computer program for determining the degree of block noise that may occur at a block boundary due to compression in a compressed image that is compressed by dividing an image into vertical and horizontal blocks of a predetermined number of pixels.

Conventionally, as this kind of block noise determination method, the quality of the image is determined based on the difference in strength across the block boundary with respect to a JPEG (Joint Photographic Experts Group) compressed image compressed every block of 8 pixels vertically and horizontally. Some have been evaluated, and others have used JPEG compressed images to evaluate image quality using measured values of frequencies at block boundaries (see Non-Patent Document 1 and Non-Patent Document 2, for example). ).
`` A Generalized Block-Edge Impairment Metric for Video Coding '', HRWu and M. Yuen, IEEE SIGNAL PROCESSING LETTERS, VOL.4, NO.11, NOVEMBER 1997 `` NO-REFERENCE PERCEPTUAL QUALITY ASSESSMENT OF COMPRESSED IMAGES '', Zhou Wang, Hamid R. Sheikh and Alan C. Bovik, Proceedings of the IEEE International Conference on Image Processing, 22-25 September 2002, Volume 1, Pages 22-25.

  However, since the above-described block noise determination method can be easily calculated, it can be evaluated efficiently, but the accuracy may be low. For example, although block noise is visually generated, the evaluation may be high.

  It is an object of the block noise determination method and program of the present invention to more appropriately determine block noise. Another object of the block noise determination method and program of the present invention is to quickly determine the degree of block noise.

  The block noise determination method and program of the present invention employ the following means in order to achieve at least a part of the above object.

The block noise determination method of the present invention includes:
A block noise determination method for determining a degree of block noise that may occur at a block boundary in accordance with compression in a compressed image obtained by dividing an image into vertical and horizontal blocks having a predetermined number of pixels and compressing the image,
(A) calculating a luminance intensity difference between each pixel based on a luminance component with respect to a pixel adjacent in a predetermined direction with respect to each pixel in the compressed image;
(B) calculating the smoothness of the luminance change in each pixel based on the difference in luminance intensity between the pixels before and after in the predetermined direction;
(C) calculating a block noise evaluation index based on a ratio of an average smoothness of luminance change in pixels located on the block boundary to an average smoothness of luminance change in pixels not located on the block boundary; ,
(D) The gist is to determine the degree of block noise based on the calculated block noise evaluation index.

  In the block noise determination method of the present invention, the luminance intensity difference between the pixels is calculated based on the luminance component for the pixels adjacent to each pixel in the compressed image in a predetermined direction, and the luminance intensity difference between the pixels is calculated. The smoothness of the luminance change in each pixel is calculated based on the difference between before and after in a predetermined direction, and the luminance of the pixel located at the block boundary with respect to the average smoothness of the luminance change in the pixel not located at the block boundary is calculated. The degree of block noise is determined by the block noise evaluation index calculated based on the average ratio of the smoothness of change. That is, since the level of block noise is determined based on the smoothness of the luminance change between the pixel located at the block boundary and the adjacent pixel, the level of block noise can be determined more appropriately. Moreover, since the calculation is performed only in accordance with the standard, the degree of block noise can be determined more quickly.

  In the block noise determination method of the present invention, the predetermined direction may be two directions of a vertical direction and a horizontal direction. In this case, the step (c) may be a step of calculating a block noise evaluation index based on an average of the ratio in the vertical direction and the ratio in the horizontal direction.

  In the block noise determination method of the present invention, the predetermined direction may be any one of a vertical direction and a horizontal direction. In this way, the degree of block noise can be determined more quickly.

  Furthermore, in the block noise determination method of the present invention, the step (a) is a step of calculating a luminance component difference with respect to a pixel adjacent to each pixel in a predetermined direction as a luminance intensity difference between the pixels. It can also be. The step (b) is a step of calculating the sum of absolute values of differences before and after the luminance intensity difference between the pixels in the predetermined direction as the smoothness of the luminance change in the pixels. You can also.

  Alternatively, in the block noise determination method of the present invention, the step (c) may be a step of calculating a block noise evaluation index so that the value becomes larger as the degree of block noise increases. In this case, the step (c) may be a step of calculating a block noise evaluation index so as to be a numerical value in the range of value 1 to value 10. In this way, the degree of block noise can be determined in the range of value 1 to value 10.

The program of the present invention
A computer program for determining the degree of block noise that may occur at a block boundary accompanying compression in a compressed image obtained by dividing an image into vertical and horizontal blocks of a predetermined number of pixels,
(A) a module that calculates a luminance intensity difference between each pixel based on a luminance component with respect to a pixel adjacent in a predetermined direction with respect to each pixel in the compressed image;
(B) a module that calculates the smoothness of the luminance change in each pixel based on the difference in luminance intensity between the pixels before and after in the predetermined direction;
(C) A module for calculating a block noise evaluation index based on a ratio of a smoothness average of luminance change in pixels located at the block boundary to a mean smoothness of luminance change in pixels not located at the block boundary When,
It is a summary to provide.

  According to the program of the present invention, the block noise evaluation index in the block noise determination method described above can be calculated by installing and starting the computer. Therefore, the luminance intensity difference between the pixels is calculated based on the luminance component for the pixels adjacent to each pixel in the compressed image in the predetermined direction, and the difference between the luminance intensity differences between the pixels in the predetermined direction is calculated. Based on this, the smoothness of the luminance change at each pixel is calculated, and the ratio of the average smoothness of the luminance change at the pixel located at the block boundary to the average smoothness of the luminance change at the pixel not located at the block boundary. The block noise evaluation index is calculated based on the block noise evaluation index, that is, the block noise evaluation index is calculated based on the smoothness of the luminance change between the pixel located at the block boundary and the adjacent pixel. Thus, the degree of block noise can be determined more appropriately.

  Next, the best mode for carrying out the present invention will be described using examples.

  In the block noise determination method of the embodiment, the degree of noise (hereinafter referred to as block noise) that can occur at a block boundary associated with compression of a JPEG (Joint Photographic Experts Group) compressed image that compresses 8 pixels vertically and horizontally as a block is determined. It will be described as being. FIG. 1 is a flowchart illustrating an example of a block noise determination method according to the embodiment. In the determination of the block noise of the embodiment, first, an image (RGB image) expressed in the color system of red (R), green (G), and blue (B) to be subjected to JPEG development is expressed by the following equation (1). ) Is converted into a YIQ color space composed of three elements of Y (luminance), I (orange-cyan), and Q (green-magenta) (step S100).

続いて、ＹＩＱ色空間に変換した画像のＹチャンネルを用いて水平方向と垂直方向における各画素間の輝度強度差を計算する（ステップＳ１１０）。実施例では、輝度強度差は、水平方向の輝度強度差をｄｘ（ｘ，ｙ）とし、垂直方向の輝度強度差をｄｙ（ｘ，ｙ）とし、各画素の輝度をＹ（ｘ，ｙ）とすると、次式（２）および式（３）により計算される。この輝度強度差は、式（２）および式（３）から解るように、隣接する画素間の輝度の偏差を表わす。輝度強度差の計算を模式的に示す模式図を図２に示す。

Subsequently, the luminance intensity difference between the pixels in the horizontal direction and the vertical direction is calculated using the Y channel of the image converted into the YIQ color space (step S110). In the embodiment, the luminance intensity difference is dx (x, y) in the horizontal direction, dy (x, y) in the vertical direction, and Y (x, y) in each pixel. Then, it is calculated by the following equations (2) and (3). This luminance intensity difference represents a luminance deviation between adjacent pixels, as can be seen from the equations (2) and (3). FIG. 2 is a schematic diagram schematically showing the calculation of the luminance intensity difference.

dx (x, y) = Y (x + 1, y) -Y (x, y) (2)
dy (x, y) = Y (x, y + 1) -Y (x, y) (3)
Next, the smoothness of the luminance change in each pixel in the horizontal direction and the vertical direction is calculated from the calculated luminance intensity difference (step S120). In the embodiment, the smoothness of the luminance change in each pixel is sx (x, y), and the smoothness of the luminance change in the vertical pixel is sy (x , Y), it is calculated by the following equations (4) and (5). The smoothness of the luminance change in this pixel represents the sum of the absolute values of the luminance deviations between adjacent pixels, as can be seen from the equations (4) and (5). FIG. 3 shows an example of the brightness of the image, the brightness intensity difference, and the smoothness of the brightness change that are schematically modeled in one dimension.

sx (x, y) = | dx (x-1, y) -dx (x, y) | + | dx (x, y) -dx (x + 1, y) | (4)
sy (x, y) = | dy (x, y-1) -dy (x, y) | + | dy (x, y) -dy (x, y + 1) | (5)
When the smoothness of the luminance change of each pixel is calculated in this way, the average ave of the smoothness of the luminance change in the horizontal direction and the vertical direction in the pixel (hereinafter referred to as the boundary pixel) located at the boundary of the block used for JPEG compression. (Psx) and ave (psy) are calculated (step S130), and the average smoothness ave (nsx) of the luminance change in the horizontal direction and the vertical direction in pixels other than the boundary pixels (hereinafter referred to as internal pixels), ave (nsy) is calculated (step S140). Specifically, the average ave (psx) of the smoothness of the luminance change at the boundary pixels in the horizontal direction is calculated by dividing the sum of sx (x, y) where x is a multiple of 8 by the number, and the vertical direction The average ave (psi) of the smoothness of the luminance change at the boundary pixels of the pixel is calculated by dividing the sum of sx (x, y), where y is a multiple of 8, by the number, and the luminance change at the horizontal internal pixel Is calculated by dividing the sum of sx (x, y) where x is not a multiple of 8 by the number of average ave (nsx) of the smoothness of the pixel, and the average ave of the smoothness change of the luminance in the vertical internal pixels (Nsy) is calculated by dividing the sum of sx (x, y) where y is not a multiple of 8 by the number.

  The ratio of the average smoothness ave (psx) of the luminance change in the horizontal boundary pixel to the average smoothness ave (nsx) of the luminance change in the horizontal inner pixel {ave (psx) / ave (nsx )} Is calculated as a block noise evaluation value Bh in the horizontal direction (step S150), and the luminance change smoothness in the vertical boundary pixels with respect to the average ave (nsy) of the luminance change smoothness in the vertical inner pixels is smoothed. The block noise evaluation value Bv in the vertical direction is calculated as a ratio of average ave (psy) {ave (psy) / ave (nsy)} (step S160), and the block noise evaluation value Bh in the horizontal direction and the block in the vertical direction are calculated. The average of the noise evaluation value Bv is calculated as the block noise evaluation value B in the image. (Step S170). Since the block noise evaluation value Bh in the horizontal direction and the block noise evaluation value Bv in the vertical direction are ratios of the smoothness of the luminance change in the boundary pixels to the average smoothness of the luminance change in the internal pixels, the degree of block noise When the value is small, the value is close to value 1, and when the degree of block noise is large, the value is larger than value 1. In addition, when the average smoothness of the luminance change in the internal pixels is small, for example, in an image where the color tone or brightness changes smoothly, the average smoothness of the luminance change in the boundary pixels is prominent. When the average of the smoothness of the luminance change in the internal pixel is large, for example, in the image in which the color tone or the brightness changes greatly, the average smoothness of the luminance change in the boundary pixel is buried. The block noise evaluation value is small. Therefore, the degree of block noise can be expressed more appropriately. Then, the degree of block noise is determined based on the block noise evaluation value B (step S180). In the embodiment, the block noise evaluation value B is 1 when there is no block noise, and close to 10 when the block noise is severe. Therefore, the degree of block noise can be determined within the range of the value 1 to 10. it can.

  According to the block noise determination method described above, the degree of block noise is determined based on the smoothness of the luminance change in the boundary pixels, and therefore the degree of block noise can be determined more appropriately. In addition, since the block noise evaluation value B is calculated as the ratio of the smoothness of the luminance change in the boundary pixel to the average smoothness of the luminance change in the internal pixel, the boundary is not used in an image in which the color tone or brightness changes smoothly. The block noise evaluation value B is calculated as a large value that emphasizes the average smoothness of the luminance change in the pixels, and the average smoothness of the luminance change in the boundary pixels is buried in the image in which the color tone and the brightness change greatly. Since the block noise evaluation value B is calculated as a small value, the evaluation based on the block noise evaluation value B can correspond to the visual evaluation. As a result, the degree of block noise can be determined more appropriately. Further, since the block noise evaluation value B in the range of the value 1 to the value 10 is used, the block noise can be determined by a numerical value.

  In the block noise determination method of the embodiment, the process of calculating the luminance intensity difference between the horizontal direction and the vertical direction in step S110 in the flowchart of FIG. 1 corresponds to step (a) in the block noise determination method of the present invention. The processing for calculating the smoothness of the luminance change in the pixels in the horizontal and vertical directions in step S120 in the flowchart corresponds to step (b) in the block noise determination method of the present invention, and steps S130 to S170 in the flowchart in FIG. The process of calculating the block noise evaluation value B corresponds to step (c) in the block noise determination method of the present invention, and the degree of block noise is determined based on the block noise evaluation value B of step S180 in the flowchart of FIG. Processing is a block of the present invention Corresponding to step (d) in the size determination method.

  In the block noise determination method of the embodiment, the block noise evaluation value B is calculated as an average of the block noise evaluation value Bh in the horizontal direction and the block noise evaluation value Bv in the vertical direction. Only the value Bh may be calculated, and the calculated block noise evaluation value Bh may be handled as the block noise evaluation value B of the image. Alternatively, only the block noise evaluation value Bv in the vertical direction is calculated, and the calculated block noise evaluation is performed. The value Bv may be handled as the block noise evaluation value B of the image. In this way, the amount of calculation for calculating the block noise evaluation value B can be reduced.

  In the block noise determination method of the embodiment, the block noise evaluation value B in the range of value 1 to value 10 is used. However, the block noise evaluation value B is adjusted to be in any range such as the range of value 1 to value 100. It may be a thing.

  In the block noise determination method of the embodiment, the smoothness of the luminance change in the pixels is calculated for all the pixels (pixels), but the smoothness of the luminance change in the pixels for all the pixels is calculated for the boundary pixels. Calculate the smoothness of the luminance change in the pixels only for the internal pixels, for example, the pixels that have a remainder of 1, 4 and the pixels that have a remainder of 2, 6 when divided by 8 for x and y, for example. It may be calculated.

  In the embodiment, it has been described that the degree of block noise that may occur at the block boundary due to compression of a JPEG compressed image that compresses the vertical and horizontal 8 pixels as a block is determined. However, block noise that may occur at the block boundary due to compression is determined. If it determines, it can apply also when compressing as a block of any size in length and breadth.

  In the embodiment, the determination method for determining the degree of block noise has been described. However, such a block noise determination method may be applied as a program form. In the case of the form of this program, it is only necessary to describe the processing shown in FIG. 1 by using an appropriate programming language so as to cause the computer to execute each processing as a procedure.

  The best mode for carrying out the present invention has been described with reference to the embodiments. However, the present invention is not limited to these embodiments, and various modifications can be made without departing from the gist of the present invention. Of course, it can be implemented in the form.

  The present invention is applicable to industries using image compression technology.

The flowchart which shows an example of the determination method of the block noise of an Example. The image figure at the time of calculating a luminance intensity difference. Explanatory drawing which shows an example of the smoothness of the change of a brightness | luminance, a brightness | luminance intensity difference, and a brightness | luminance.

Claims (9)

A block noise determination method for determining a degree of block noise that may occur at a block boundary in accordance with compression in a compressed image obtained by dividing an image into vertical and horizontal blocks having a predetermined number of pixels and compressing the image,
(A) calculating a luminance intensity difference between each pixel based on a luminance component with respect to a pixel adjacent in a predetermined direction with respect to each pixel in the compressed image;
(B) calculating the smoothness of the luminance change in each pixel based on the difference in luminance intensity between the pixels before and after in the predetermined direction;
(C) calculating a block noise evaluation index based on a ratio of an average smoothness of luminance change in pixels located on the block boundary to an average smoothness of luminance change in pixels not located on the block boundary; ,
(D) A block noise determination method for determining the degree of block noise based on the calculated block noise evaluation index.   The block noise determination method according to claim 1, wherein the predetermined direction is two directions of a vertical direction and a horizontal direction.   3. The block noise determination method according to claim 2, wherein the step (c) is a step of calculating a block noise evaluation index based on an average of the ratio in the vertical direction and the ratio in the horizontal direction.   The block noise determination method according to claim 1, wherein the predetermined direction is one of a vertical direction and a horizontal direction.   5. The block noise determination according to claim 1, wherein the step (a) is a step of calculating a difference in luminance component with respect to a pixel adjacent to each pixel in a predetermined direction as a luminance intensity difference between the pixels. Method.   6. The step (b) is a step of calculating a sum of absolute values of differences before and after the luminance intensity difference between pixels in the predetermined direction as smoothness of luminance change in each pixel. One of the block noise determination methods.   The block noise determination method according to claim 1, wherein the step (c) is a step of calculating a block noise evaluation index so that the block noise evaluation index becomes larger as the degree of block noise increases.   8. The block noise determination method according to claim 7, wherein the step (c) is a step of calculating a block noise evaluation index so as to be a numerical value in a range of value 1 to value 10. A program for a computer for determining the degree of block noise that may occur at a block boundary due to compression in a compressed image obtained by dividing an image into vertical and horizontal blocks of a predetermined number of pixels and compressing the image,
(A) a module that calculates a luminance intensity difference between each pixel based on a luminance component with respect to a pixel adjacent in a predetermined direction with respect to each pixel in the compressed image;
(B) a module for calculating the smoothness of the luminance change in each pixel based on the difference between the luminance intensity differences between the pixels in the predetermined direction before and after,
(C) A module for calculating a block noise evaluation index based on a ratio of a smoothness average of luminance change in pixels located at the block boundary to a mean smoothness of luminance change in pixels not located at the block boundary When,
A program comprising

Images