JP2004159311A - Image processing apparatus and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus in which a picture quality is prevented from being extremely changed by presence/non-presence of edge by suppressing blurring of both an edge and a texture without damaging a mosquito noise elimination effect. <P>SOLUTION: An image processing apparatus 1 is equipped with: a differential value calculating part 11 for calculating a differential value between a pixel value of a concerned pixel in a filter processing region set for each of pixels comprising a decoded image and pixel values of peripheral pixels; a distribution coefficient calculating part 12 for calculating distribution coefficients of pixel values in the pixels; a filter coefficient calculating part 13 for calculating a filter coefficient with respect to the peripheral pixels based upon the calculated differential value and distribution coefficient; and a filter processing part 14 for applying filter processing to the pixel value of the concerned pixel in the decoded image while using the filter coefficient calculated by the filter coefficient calculating part 13, to calculate the pixel value of the concerned pixel in the decoded image. <P>COPYRIGHT: (C)2004,JPO

Description

  The present invention provides an image processing apparatus for generating a restored image in which mosquito noise generated by block coding is effectively removed by applying a filtering process to a decoded image without blurring edges and textures in the image. And an image processing method.

  2. Description of the Related Art In recent years, digital still cameras (Digital Still Cameras) have become widespread, and the use of digital still cameras to capture images (still images) of landscapes, people, and the like is becoming common. However, the data amount of a digitized image is very large. When an image having a large data amount is stored as it is in a storage medium such as an IC memory or transmitted through a transmission medium such as the Internet or a LAN, only a small number of images can be stored in the storage medium, Time will be longer. For this reason, a compression technique is essential for handling images.

  When image data needs to be highly compressed, an irreversible image compression method in which the original image and the decoded image do not completely match is generally used. Many irreversible image compression methods employ a method in which image data is divided into blocks of M × N pixels, and orthogonal transformation is performed in each block, and then orthogonal transformation coefficients are quantized and encoded. One of the typical methods is JPEG, which is widely used as a color still image compression method.

FIG. 15 is a block diagram illustrating a functional configuration of a JPEG encoding device and a JPEG decoding device.
The encoding device 60 compresses an original image into JPEG compressed data, and as shown in FIG. 15, a preprocessing unit 61, a DCT transform unit 62, a quantization unit 63, and a quantization table 64. And an entropy encoding unit 65. The decoding device 70 expands the JPEG compressed data into a decoded image. As shown in FIG. 15, the entropy decoding unit 71, the inverse quantization unit 72, the inverse quantization table 73, and the A DCT conversion unit 74 and a post-processing unit 75 are provided.

When performing the JPEG encoding process in the encoding device 60, first, the pre-processing unit 61 performs processing for each pixel of the original image composed of multi-valued data of Red (R), Green (G), and Blue (B). The data is converted into data of a luminance component (Y) and color difference components (Cr, Cb).
Next, the DCT transform unit 62 performs a discrete cosine transform (DCT) on the YCbCr data for each block of 8 × 8 pixels to calculate a DCT coefficient.

Next, the quantization unit 63 performs quantization of the DCT coefficient. At this time, each component of the DCT coefficient is quantized with a different step width according to the quantization table 64.
Lastly, the entropy encoding unit 65 encodes the quantized DCT coefficients to generate JPEG compressed data. In the standard method of JPEG, Huffman coding is used as entropy coding.

The above processing is the outline of the encoding processing from image data to JPEG compressed data. The JPEG compressed data generated in this way is passed to the decoding device 70 via a storage medium (for example, an SD card) or a transmission medium.
Next, a procedure of a decoding process from JPEG compressed data to a decoded image will be described.
When the decoding device 70 performs the JPEG decoding process, first, the entropy decoding unit 71 performs entropy decoding on the JPEG compressed data.

Next, the inverse quantization unit 72 performs inverse quantization. At this time, information of the quantization table 64 used at the time of encoding is read from the JPEG compressed data and used as the inverse quantization table 73.
Next, the inverse DCT unit 74 performs an inverse discrete cosine transform (IDCT) to convert the DCT coefficients into a decoded image of YCbCr data.
Finally, the post-processing unit 75 obtains a decoded image by performing a conversion process from YCbCr data to RGB data.

The above is the outline of the encoding process and the decoding process regarding JPEG.
As described above, the process of JPEG encoding includes quantization of DCT coefficients. For this reason, data degradation occurs due to the quantization error. As a result, if the decoded image is directly reproduced on paper by a printer or the like, this deterioration appears as noise in the decoded image. Even in the case of moving image compression using block coding, these noises are annoying, but in the case of a still image, particularly detailed noise can be noticeable because it can be seen carefully.

  Among the noises that occur in the decoded image, those that have a visual adverse effect include noise called mosquito noise. The mosquito noise refers to fluctuation of a weak gradation that occurs around an edge of a decoded image. This is because strong edges that existed in the original image were not accurately reproduced due to the fact that many of the high-frequency components were lost due to the quantization of the DCT coefficients.

  In order to remove such noise, as shown in FIG. 16, an image processing apparatus 100 that performs a filtering process on a decoded image to generate a restored image from which mosquito noise has been removed has been conventionally considered. A number of filtering processes have been proposed as a filtering method. One of the filtering processes is a new filtering process capable of suppressing blurring of edges and removing noise. There is a filter (for example, see Non-Patent Document 1). Note that this technology is also referred to as “first conventional technology” below.

である。 However,

It is.

Here, in Expressions (1) to (4), (x, y) represents the position of the target pixel, (i, j) represents the relative position of each peripheral pixel with respect to the target pixel, and f (X, y) represents the pixel value of the pixel of interest in the decoded image, f (x + i, y + j) represents the pixel value of each peripheral pixel in the decoded image, and g (x, y) after filtering. Represents the pixel value of the pixel of interest. Α _{x, y} (i, j) represents a filter coefficient of each peripheral pixel (i, j) with respect to the target pixel (x, y), and the first filter coefficient α1 _{x, y} (i, j) And the second filter coefficient α2 _{x, y} (i, j). Here, the first filter coefficient α1 _{x, y} (i, j) is designed such that the filter coefficient becomes larger as the peripheral pixel is closer to the target pixel, and the second filter coefficient α2 _{x, y} (i , J) are designed such that the filter coefficient increases as the peripheral pixel has a pixel value closer to the pixel value of the target pixel.

Here, σ is a parameter for adjusting the influence of the distance from the target pixel to each peripheral pixel on the first filter coefficient α1 _{x, y} (i, j), and t is the difference between the pixel value of the target pixel and the peripheral pixel. This parameter adjusts the effect of the value on the second filter coefficient α2 _{x, y} (i, j).
The greatest feature of this filter lies in the method of calculating the second filter coefficient α2 _{x, y} (i, j).

Here, FIG. 17 shows the relationship between the difference value between the pixel value of the target pixel and the pixel value of each peripheral pixel, and the value of the second filter coefficient α2 _{x, y} (i, j) calculated from the difference value. Show.
As can be seen from FIG. 17, when the pixel value of the peripheral pixel is close to the pixel value of the target pixel, a large value is set to the second filter coefficient α2 _{x, y} (i, j), and conversely, Is significantly different from the pixel value of the target pixel, a small value is set to the second filter coefficient α2 _{x, y} (i, j).

  Here, when a large value is set for the parameter t, a large filter coefficient is set for peripheral pixels having a pixel value relatively different from the pixel value of the target pixel, which corresponds to strong smoothing processing. ing. Conversely, when a small value is set for the parameter t, a small filter coefficient is set for peripheral pixels having a pixel value very close to the pixel value of the target pixel. I have. Therefore, when the SUSAN filter is applied, it is necessary to set the parameter t to an optimum constant value in advance according to the purpose of the processing.

Consider a case in which mosquito noise is removed using this SUSAN filter.
As described above, mosquito noise occurs as a fluctuation of a weak gradation around a strong edge.
FIG. 18 is a diagram schematically illustrating a relationship between an edge and mosquito noise. It is assumed that the value of the parameter t of the SUSAN filter is set sufficiently smaller than the gradation change of the edge and sufficiently larger than the gradation change of the mosquito noise. Here, when the SUSAN filter is applied to the pixel in the mosquito noise portion, the pixel in the edge portion having a pixel value greatly different from the pixel value of the target pixel due to the effect of the second filter coefficient α2 _{x, y} (i, j). , A very small filter coefficient is set, so that there is substantially no effect on the filtering process, and a pixel in the mosquito noise portion near the target pixel having a pixel value close to the pixel value of the target pixel is set. Only when a large filter coefficient is set, filter processing is performed. Therefore, mosquito noise can be effectively removed without blurring edges.

  From the above considerations, in order to remove the mosquito noise while maintaining the sharpness of the edge, it is very effective to use the filter processing for determining the filter coefficient based on the difference between the pixel value of the target pixel and the pixel values of the peripheral pixels. It is considered to be.

  An image processing apparatus for removing noise for each block from image data divided into a plurality of blocks, wherein a frequency distribution of a difference value between adjacent pixels for each block based on the image data for each block And a selection unit that selects a block from which noise is to be removed based on a result determined by the determination unit based on a frequency distribution. 2. Description of the Related Art An image processing apparatus has been conventionally considered (see Patent Document 1). Note that this technology is also referred to as “second conventional technology” below.

According to the second related art, as shown in FIG. 19, a filter (ε filter) process is performed on a block including an edge, and a filter process is not performed on a block including no edge. In the processing, unnecessary mosquitoes can be removed, and blurring of the texture can be prevented.
SM Smith and JM Brady, "SUSAN-A New Approach to Low Level Image Processing," International Journal of Computer Vision, vol.23, no.1, pp.45-78, 1997. Japanese Patent No. 30111828 (page 1, FIG. 1)

  However, according to the first conventional technique, when a SUSAN filter is used to remove mosquito noise, as shown in FIG. 20, the edge is not blurred even if mosquito noise generated in the signboard area is removed. Although there is a remarkable effect, there is a problem that a very large blur occurs in a texture region such as a leaf of a tree. This is because, in order to effectively remove mosquito noise, it is necessary to set the value of the parameter t to a value sufficiently larger than the magnitude of the gradation change of the mosquito noise. This is because when a texture region is filtered, even a gradation change of the existing texture from the original image is removed.

On the other hand, according to the second conventional technique, although blurring of the texture can be suppressed without impairing the mosquito noise removal effect, the mosquito noise remains because the filtering processing is not performed on the weak edge, or the mosquito noise remains. There is a new problem that the image quality changes drastically without it.
Therefore, in the present invention, blurring of both edges and textures is suppressed without impairing the mosquito noise removal effect, and even on weak edges, mosquito noise remains or the image quality becomes extremely poor without edges. It is an object of the present invention to provide an image processing apparatus and an image processing method that prevent the image processing from being changed.

  In order to solve the above problems, in the image processing apparatus according to the present invention, a multi-valued image is decoded into a decoded image obtained by decoding compressed data obtained by encoding each block of M × N pixels. An image processing apparatus for generating a restored image by performing a filter process for removing noise, wherein a pixel value of a pixel of interest in a filter processing region set for each of the pixels constituting the decoded image. A difference value calculation unit that calculates a difference value between the pixel value of the peripheral pixel, a distribution coefficient calculation unit that calculates a distribution coefficient of a pixel value of each of the pixels, and the difference value calculated by the difference value calculation unit. Filter coefficient calculating means for calculating a filter coefficient for the peripheral pixel based on the distribution coefficient calculated by the distribution coefficient calculating means, and the filter coefficient calculating means Filter processing means for performing filter processing on the pixel value of the pixel of interest in the decoded image using the calculated filter coefficient, and calculating the pixel value of the pixel of interest in the restored image. .

  Here, the larger the distribution coefficient, the larger the filter coefficient within a range that does not affect the edge. By doing so, in the filter processing region including the edge, since the value of the distribution coefficient is large, a large filter coefficient is set for peripheral pixels having a pixel value relatively different from the pixel value of the target pixel. Strong smoothing is performed, and mosquito noise can be effectively removed without blurring edges. On the other hand, the smaller the distribution coefficient, the smaller the filter coefficient. In this way, in a filter processing area such as a texture that does not include an edge, since the value of the distribution coefficient is small, a small filter coefficient is set to a peripheral pixel having a pixel value very close to the pixel value of the target pixel. Then, weak smoothing is performed, and occurrence of blur can be suppressed. In addition, since appropriate filtering is performed on all pixels instead of processing that separates blocks into blocks that do not perform filtering and blocks that do not perform filtering as in the past, mosquito noise remains or includes edges. It is possible to reliably prevent a situation in which the image quality is extremely changed between the block and the block not included.

Specifically, the image processing apparatus according to the present invention may be configured such that the distribution coefficient calculation means calculates a distribution coefficient of a pixel value of each of the pixels for each of the blocks.
This makes it possible to easily calculate the distribution coefficient of the pixel value for each of the blocks, and to perform an appropriate filtering process on all the pixels.

Further, in the image processing device according to the present invention, the distribution coefficient calculating means calculates a distribution coefficient of a pixel value for each block, and calculates a distribution coefficient of a pixel value in the calculated block and a surrounding coefficient adjacent to the block. Calculating a pixel value interpolation distribution coefficient for each pixel by interpolating the pixel value distribution coefficient for each block, and using the calculated interpolation distribution coefficient as a pixel value distribution coefficient for each pixel. You can also.
As a result, since the distribution coefficient changes continuously even at the block boundary, it is possible to realize a filtering process in which the smoothing strength changes continuously.

  Further, in the image processing device according to the present invention, the distribution coefficient calculation unit calculates the edge strength of the peripheral pixels in a predetermined edge influence calculation area, and the edge strength calculation unit calculates the edge strength. Calculating an edge influence on the target pixel based on the edge strength of the peripheral pixel and the distance between the target pixel and the peripheral pixel, and calculating the maximum value of the calculated edge influence on the distribution coefficient of the pixel value of each pixel And a maximum edge influence degree calculating means that calculates the degree of influence.

As a result, even when the block boundary is unknown, the distribution coefficient changes continuously based on the edge strength of the peripheral pixel and the distance between the pixel of interest and the peripheral pixel. Can be realized.
Here, the distribution coefficient calculation means may calculate a difference value obtained by subtracting a minimum value from a maximum value of pixel values in the block as a distribution coefficient of pixel values of the block.

Thus, the distribution coefficient can be easily obtained.
Further, the distribution coefficient calculating means may calculate a variance value from an average pixel value in the block as a distribution coefficient of a pixel value of the block.
This also allows the distribution coefficient to be easily obtained.

Further, the distribution coefficient calculation means may calculate a maximum edge intensity in the block as a distribution coefficient of a pixel value of the block.
This also allows the distribution coefficient to be easily obtained.

  Note that the present invention can be realized not only as such an image processing apparatus, but also as an image processing method in which the characteristic means of such an image processing apparatus are implemented as steps. As a program to be executed. Needless to say, such a program can be distributed via a recording medium such as a CD-ROM or a transmission medium such as the Internet.

  As is apparent from the above description, according to the image processing apparatus of the present invention, in the filter processing area including the edge, the distribution coefficient has a large value, and thus has a pixel value that is relatively significantly different from the pixel value of the target pixel. A large filter coefficient is set for peripheral pixels, and strong smoothing is performed, so that mosquito noise can be effectively removed. On the other hand, in a filter processing area such as a texture that does not include an edge, since the value of the distribution coefficient is small, a small filter coefficient is set to a peripheral pixel having a pixel value very close to the pixel value of the target pixel, and weak smoothing is performed. Is performed, and the occurrence of blur can be suppressed. In addition, since appropriate filtering is performed on all pixels instead of processing that separates blocks into blocks that do not perform filtering and blocks that do not perform filtering as in the past, mosquito noise remains or includes edges. It is possible to reliably prevent a situation in which the image quality is extremely changed between the block and the block not included.

  Therefore, according to the present invention, it is possible to reproduce a high-quality restored image, and the practical value of the present invention is extremely high today, in addition to digital cameras and the like, as well as high-definition printers and projectors.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(Embodiment 1)
FIG. 1 is a diagram showing an external configuration of a digital camera and a printer.
The digital camera ex100 includes a color image sensor such as a CCD or a CMOS, and the above-described image encoding device 60 (see FIG. 15), and blocks the original image acquired by the color image sensor by the encoding device 60. By encoding each time, JPEG compressed data is generated, and the generated JPEG compressed data is recorded on the SD card ex200.

The printer ex400 has a so-called direct print function, and includes, in addition to the card reader ex301 for mounting the SD card ex200 and the above-described image decoding device 70, an image processing device, a preprocessing unit, a printer engine, and the like. Be composed.
The printer ex400 reads the JPEG compressed data specified by the user from the SD card ex200 and decodes the data by the entropy decoding unit 71, the inverse quantization unit 72, the inverse quantization table 73, and the inverse DCT transform unit 74 of the decoding device 70. , YCbCr data. Note that the decoded image contains mosquito noise generated near the edge.

  The image processing apparatus removes mosquito noise, suppresses edge and texture blur by performing a filtering process on the Y component in the decoded image of the YCbCr data output from the inverse DCT transform unit 74, and furthermore, A restored image is generated that prevents mosquito noise from remaining even on a weak edge and prevents the image quality from extremely changing without an edge.

The post-processing unit 75 of the decoding device 70 converts the Y component filtered by the image processing device and the original CbCr data into RGB data of a restored image.
The preprocessing unit converts the RGB color restored image into a YMCK color restored image.
The printer engine reproduces a YMCK color restored image on a sheet. Here, in the restored image generated by the image processing device, mosquito noise is removed while suppressing blurring of edges and textures and preventing the image quality from being extremely changed without an edge. I have. Therefore, even in the restored image reproduced on the paper, as shown in FIG. 1, while suppressing the blur of the edge and the texture, and preventing the image quality from extremely changing without the edge, Mosquito noise can be removed regardless of whether the edge strength is high or low.

FIG. 2 is a block diagram showing a functional configuration of the image processing apparatus mounted on the printer ex400 shown in FIG.
The image processing apparatus 1 generates a restored image by performing a filtering process for each block of the Y component of the decoded image, and as shown in FIG. It includes a coefficient calculator 12, a filter coefficient calculator 13, and a filter processor 14.

The difference value calculation unit 11 calculates a difference value between the pixel value of the target pixel and the pixel values of each peripheral pixel.
The distribution coefficient calculation unit 12 calculates a distribution coefficient of a pixel value in each block.
The filter coefficient calculation unit 13 calculates a filter coefficient for each peripheral pixel based on the difference value calculated by the difference value calculation unit 11 and the distribution coefficient calculated by the distribution coefficient calculation unit 12.

The filter processing unit 14 performs a filter process using the filter coefficients for each peripheral pixel calculated by the filter coefficient calculation unit 13, and generates a restored image from which noise has been removed.
The difference value calculation unit 11, the distribution coefficient calculation unit 12, the filter coefficient calculation unit 13, and the filter processing unit 14 included in the image processing apparatus 1 are a CPU, a ROM in which an image processing program is stored in advance, and an image processing program. It is configured by a RAM or the like that provides a work area or the like for execution.

The filter processing unit 14 of the image processing apparatus 1 can be represented by the following equation (5).

である。 However,

It is.

Here, in the above equations (5) to (8), (x, y) represents the position of the pixel of interest. (I, j) indicates the relative position of each peripheral pixel to the target pixel. f (x, y) represents the pixel value of the pixel of interest in the decoded image. f (x + i, y + j) represents the pixel value of each peripheral pixel in the decoded image. g (x, y) represents the pixel value of the target pixel in the restored image after the filter processing. (X, Y) represents the position of the block including the target pixel (x, y). β _{x, y} (i, j) represents a filter coefficient of each peripheral pixel (i, j) with respect to the target pixel (x, y). ν _{x, y} (i, j) represents a difference value between the pixel value f (x, y) of the target pixel and the pixel value f (x + i, y + j) of each peripheral pixel. U (X, Y) represents a value (distribution coefficient) obtained by subtracting the minimum value from the maximum value of the pixel values in the block (X, Y). K represents a constant value set in advance.

Next, an operation when the image processing apparatus 1 configured as described above generates a restored image from a decoded image will be described.
FIG. 3 is a flowchart illustrating an operation of the image processing apparatus 1 when generating a restored image from a decoded image. When generating the restored image, the image processing apparatus 1 previously acquires information on where the block boundary line exists in the decoded image in the process of decoding the JPEG compressed data.

  The image processing apparatus 1 repeatedly executes the following processing for each block (S11). First, the distribution coefficient calculation unit 12 sequentially obtains the maximum value and the minimum value of the pixel value for each block of M × N pixels (the area divided by the block boundary shown in FIG. 4) (S12). , S13), the minimum pixel value min {f (x, y)} is subtracted from the maximum pixel value max {f (x, y)} in the block, and the resulting value is used as the distribution coefficient U ( X, Y) (S14).

When the calculation of the distribution coefficient applied to each block is completed (S15), the difference value calculation unit 11 repeatedly executes the following processing for each pixel of the decoded image (S16). The difference value calculation unit 11 calculates the pixel value f (x, y) of a certain target pixel and the pixel value f (x, y) of each peripheral pixel in the filter processing area (all the peripheral pixels included in the filter processing area shown in FIG. 4). x + i, y + j) and the difference values ν _{x, y} (i, j) are calculated (S17).

  Here, FIG. 4 shows a positional relationship between an area of a pixel of interest, peripheral pixels used for filter processing (filter processing area), and a block divided into M × N pixels by a block boundary line. Although FIG. 4 shows a case where the filter size is 5 × 5 and the block size is 8 × 8, the present invention is not limited to this size. Further, the filter processing area is not limited to a rectangle.

When the calculation of all the difference values v _{x, y} (i, j) is completed (S18), the filter coefficient calculation unit 13 repeatedly executes the following processing for each peripheral pixel in the filter processing area (S19). The filter coefficient calculating unit 13 calculates each of the difference values ν _{x, y} (i, j) calculated by the difference value calculating unit 11 and the distribution coefficient U (X, Y) calculated by the distribution coefficient calculating unit 12. A filter coefficient β _{x, y} (i, j) for peripheral pixels is calculated (S20). Specifically, the filter coefficient β _{x, y} (i, j) can be obtained by the calculation expression shown in Expression (6). Here, for the value of the distribution coefficient U (X, Y), a value calculated in the block including the target pixel is used, and for the difference value ν _{x, y} (i, j), Use the calculated values.

The constant K in the equation (6) is a parameter for adjusting the influence of the distribution coefficient U (X, Y) on the calculation of the filter coefficient β _{x, y} (i, j). Can be set.
When all the filter coefficients β _{x, y} (i, j) have been calculated (S21), the filter processing unit 14 repeatedly executes the following processing for each pixel of interest (S22). The filter processing unit 14 performs filter processing on the pixel of interest using the filter coefficient β _{x, y} (i, j) for each peripheral pixel calculated by the filter coefficient calculation unit 13, and performs pixel processing on the pixel of interest after the filter processing. A value is calculated (S23). When such processing (S23) is repeatedly executed and the pixel values after the filter processing are calculated for each pixel in all the blocks (S24), a restored image is generated, and the restored image generation processing ends.

  When the image processing apparatus 1 according to the first embodiment is used, in a block including an edge, the value of the distribution coefficient U (X, Y) is large, so that a peripheral pixel having a pixel value relatively different from the pixel value of the target pixel is used. A large filter coefficient is set for the pixel within a range that does not affect the edge, strong smoothing is performed, and mosquito noise can be effectively removed. On the other hand, in a block such as a texture that does not include an edge, since the value of the distribution coefficient U (X, Y) is small, a small filter coefficient is set for a peripheral pixel having a pixel value very close to the pixel value of the target pixel. Weak smoothing is performed, and the occurrence of blur can be suppressed. Moreover, since appropriate filtering is performed on all blocks, instead of processing that separates blocks into blocks that do not perform filtering and blocks that do not perform filtering, mosquito noise remains and edges are included. It is possible to reliably prevent a situation in which the image quality is extremely changed between the block and the block not included.

(Embodiment 2)
FIG. 5 is a block diagram illustrating another functional configuration of the image processing apparatus mounted on the printer ex400.
The image processing device 2 generates a restored image by performing a filtering process for each block of the Y component of the decoded image. As shown in FIG. It comprises a coefficient calculation unit 12, a distribution coefficient interpolation unit 22, a filter coefficient calculation unit 23, and a filter processing unit 24. Note that the same reference numerals are given to portions corresponding to the configuration of the image processing apparatus 1, and description thereof will be omitted.

  Here, in the image processing apparatus 1 according to the first embodiment, the filter coefficient is calculated using the same distribution coefficient U (X, Y) (uniform value) in each block. Generally, in the block coding, since the coding process is performed independently for each block, noise such as mosquito noise is also generated for each block. Therefore, by adjusting the level of smoothing in block units as in the above processing, blurring of the texture can be suppressed, and mosquito noise can be effectively removed.However, depending on the image, In some cases, it may be preferable to perform filter processing so that the smoothing strength does not become discontinuous at the block boundary.

  Therefore, in the image processing apparatus 2, a plurality of distribution coefficients of the distribution coefficient of the block to be processed and the distribution coefficient of the peripheral block are used, and the distribution coefficient of the block and the distribution coefficient of the peripheral block are determined in the two-dimensional direction. It further includes a distribution coefficient interpolation unit 22 for calculating an interpolation distribution coefficient u (x, y) that continuously changes for each pixel by performing linear interpolation, and calculating a filter coefficient using this value. It is configured to realize a filtering process in which the smoothing strength changes continuously even at the block boundary.

Next, an operation when the image processing apparatus 2 configured as described above generates a restored image from a decoded image will be described.
FIG. 6 is a flowchart illustrating an operation of the image processing apparatus 2 when generating a restored image from a decoded image. Note that, similarly to the image processing apparatus 1, when generating the restored image, the image processing apparatus 2 previously acquires information on where the block boundary line exists in the decoded image in the process of decoding the JPEG compressed data. I have.

  The image processing device 2 repeatedly executes the following processing for each block (S11). First, the distribution coefficient calculation unit 12 sequentially obtains the maximum value and the minimum value of the pixel value for each block of M × N pixels (the area divided by the block boundary shown in FIG. 4) (S12). , S13), the minimum pixel value min {f (x, y)} is subtracted from the maximum pixel value max {f (x, y)} in the block, and the resulting value is used as the distribution coefficient U ( X, Y) (S14).

When the calculation of the distribution coefficient applied to each block is completed (S15), the difference value calculation unit 11 repeatedly executes the following processing for each pixel of the decoded image (S16). The difference value calculation unit 11 calculates the pixel value f (x, y) of a certain target pixel and the pixel value f (x, y) of each peripheral pixel in the filter processing area (all the peripheral pixels included in the filter processing area shown in FIG. 4). x + i, y + j) and the difference values ν _{x, y} (i, j) are calculated (S17).

When the calculation of all the difference values v _{x, y} (i, j) is completed (S18), the distribution coefficient interpolation unit 22 repeatedly executes the following processing for each pixel (S31). The distribution coefficient interpolation unit 22 calculates the distribution coefficient U (X, Y) of the block of the pixel of interest and the distribution coefficients U (X-1, Y-1) and U (X, Y-1) of the peripheral blocks shown in FIG. , U (X-1, Y), an interpolation distribution coefficient u (x, y) of a pixel value at each pixel is calculated (S32).

This interpolation method will be described with reference to FIG.
First, it is assumed that the distribution coefficient U (X, Y) calculated by the distribution coefficient calculation unit 12 is the distribution coefficient at the center point of each block.

  Next, when the center point of each block and the center points of four adjacent blocks are connected by a straight line, each pixel of the decoded image is included in any of the rectangular regions surrounded by the center points of the four blocks. Here, the distribution coefficients of the center points of four blocks surrounding a certain pixel of interest (x, y) are U (X, Y), U (X-1, Y), U (X, Y-1), If U (X-1, Y-1), the interpolation distribution coefficient u (x, y) for the pixel of interest (x, y) is obtained by dividing these four distribution coefficients from the center point to the pixel of interest (x, y). It can be calculated by performing linear interpolation according to the distance to y).

When the calculation of all the interpolation distribution coefficients u (x, y) is completed (S33), the filter coefficient calculation unit 23 repeatedly executes the following processing for each peripheral pixel (S19). The filter coefficient calculation unit 23 calculates the difference value ν _{x, y} (i, j) calculated by the difference value calculation unit 11 and the interpolation distribution coefficient u (x, y) calculated by the distribution coefficient interpolation unit 22. A filter coefficient β _{x, y} (i, j) for peripheral pixels is calculated (S20). Specifically, a filter coefficient is calculated using an expression obtained by replacing the distribution coefficient U (X, Y) in Expression (8) with the interpolation distribution coefficient u (x, y) calculated for the position of each pixel of interest. β _{x, y} (i, j) is calculated.

When all the filter coefficients β _{x, y} (i, j) have been calculated (S21), the filter processing unit 24 repeatedly executes the following processing for each pixel of interest (S22). The filter processing unit 24 performs filter processing on the target pixel using the filter coefficient β _{x, y} (i, j) for each peripheral pixel calculated by the filter coefficient calculation unit 23, and performs pixel processing on the target pixel after the filter processing. A value is calculated (S23). When such a process (S23) is repeatedly executed and the pixel value after the filter process is calculated for each target pixel in all the blocks (S24), a restored image is generated, and the restored image generation process ends.

  When the image processing apparatus 2 according to the second embodiment is used, not only the same effect as the image processing apparatus 1 can be obtained, but also the continuous processing is performed for each pixel calculated by interpolating the distribution coefficient U (X, Y). Since the filter coefficient is calculated using the interpolation distribution coefficient u (x, y), the smoothing strength does not become discontinuous even at the block boundary, and the smoothing strength changes continuously. Filter processing can be realized.

(Embodiment 3)
FIG. 8 is a diagram showing an external configuration of a digital camera, a personal computer (hereinafter, also referred to as “PC”), a magnification converter, and a projector. Note that the digital camera ex100 and the SD card ex200 are the same as those in FIG.

  The PCex300 includes a card reader ex301 for mounting the SD card ex200, the above-described image decoding device 70, the display ex302, and the like. The PCex300 reads JPEG compressed data specified by a user from the SD card ex200 and reads the JPEG compressed data from the decoding device. 70 to generate an RGB color decoded image. Then, the PCex300 displays the generated decoded image on the display ex302 or outputs it to the magnification converter ex500. Note that the decoded image displayed on the display ex302 includes mosquito noise generated near the edge.

The magnification converter ex500 adjusts the size of the decoded RGB color image output from the PCex300, and outputs the decoded RGB color decoded image. When the size is reduced or enlarged, a thinning process or an interpolation process is appropriately performed.
The projector ex600 includes a pre-processing unit, an image processing device, a post-processing unit, a display engine, and the like.

The preprocessing unit converts the size-adjusted decoded image of RGB color output from the magnification converter ex400 into a decoded image of YCbCr data.
The image processing apparatus removes mosquito noise, suppresses blurring of edges and textures by applying a filtering process to the Y component of the decoded image of the YCbCr data output from the preprocessing unit, and furthermore, controls weak edges. Also, a restored image is generated in which mosquito noise is not left and the image quality is prevented from being extremely changed without an edge.

The post-processing unit converts the Y component filtered by the image processing device and the original CbCr data into RGB data of a restored image.
The display engine focuses the restored RGB color image on the screen ex700 and reproduces the restored RGB color image. Here, in the restored image generated by the image processing device, mosquito noise is removed while suppressing blurring of edges and textures and preventing image quality from extremely changing without an edge. . Therefore, even in the restored image reproduced on the screen ex700, as shown in FIG. 8, blur of edges and textures is suppressed, and the image quality is prevented from being extremely changed without an edge. The mosquito noise can be removed regardless of whether the edge strength is high or low.

Next, an image processing device mounted on the projector ex600 will be described.
FIG. 9 is a block diagram illustrating a functional configuration of an image processing device mounted on projector ex600 shown in FIG.
The image processing device 3 generates a restored image by performing a filtering process on the Y component of each pixel constituting the decoded image for each predetermined filtering process region, as shown in FIG. , A difference value calculation unit 11, an edge strength calculation unit 31, a maximum edge influence calculation unit 32, a filter coefficient calculation unit 33, and a filter processing unit 34. Parts corresponding to the components of the image processing apparatuses 1 and 2 are denoted by the same reference numerals, and description thereof is omitted.

  By the way, the image processing apparatus 1 calculates a filter coefficient using the same distribution coefficient U (X, Y) in each block. Further, the image processing apparatus 2 calculates the interpolated distribution coefficient u (x, y) which continuously changes for each pixel by interpolating the distribution coefficient U (X, Y) calculated for each block, and calculates this value. By calculating the filter coefficients using the filter coefficients, filter processing in which the smoothing strength continuously changes even at the block boundary is realized. However, when the decoded image is enlarged or reduced as described above, it is not possible to perform an appropriate filtering process because it is not known where the block boundary exists. Therefore, the image processing device 3 is configured to perform appropriate filtering processing that does not depend on the block boundaries even if the block boundaries are not known.

  The edge strength calculation unit 31 of the image processing device 3 is composed of, for example, a Sobel filter (“Image Analysis Handbook”, supervised by Mikio Takagi and Hirohisa Shimoda, 550-564, University of Tokyo Press, 1991), and focuses on the pixel of interest. The edge intensity is calculated for each peripheral pixel in the edge influence calculation area (for example, 11 × 11 pixels), and the edge intensity of each peripheral pixel is detected.

The maximum edge influence calculation unit 32 calculates the edge influence of each peripheral pixel based on the edge strength of the peripheral pixel in the edge influence calculation area and the distance between the pixel of interest and the peripheral pixel. Then, the maximum edge influence calculation unit 32 determines the maximum value of the edge influences of all the peripheral pixels as the distribution coefficient used in the filter processing area.
The filter coefficient calculation unit 33 calculates a difference value between the pixel value of the peripheral pixel and the pixel value of the target pixel in the filter processing area calculated by the difference value calculation unit 11 and the filter calculated by the maximum edge influence degree calculation unit 32. Based on the distribution coefficients used in the processing area, filter coefficients of peripheral pixels in the filtering area are calculated.

The filter processing unit 34 performs filter processing on the pixel of interest using the filter coefficient β _{x, y} (i, j) for each peripheral pixel calculated by the filter coefficient calculation unit 33, and performs filtering on the pixel of interest after the filter processing. Calculate the value. As a result, the filter processing unit 34 performs strong smoothing within a range that does not affect the edge as the edge is closer to a strong edge, and weaker as the distance from the edge is greater. That is, the filter processing unit 34 dynamically changes the filter coefficient depending on the edge strength of the peripheral pixels and the distance.

The difference value calculation unit 11, the edge strength calculation unit 31, the maximum edge influence degree calculation unit 32, the filter coefficient calculation unit 33, and the filter processing unit 34 included in the image processing device 3 are the same as those of the image processing devices 1 and 2. In the same manner as described above, it is constituted by a CPU, a ROM storing an image processing program in advance, a RAM providing a work area for executing the image processing program, and the like.
Next, an operation when the image processing apparatus 3 configured as described above generates a restored image from a decoded image will be described.

FIG. 10 is a flowchart illustrating an operation of the image processing apparatus 3 when generating a restored image from a decoded image.
The image processing device 3 repeatedly executes the following processing for each edge influence degree calculation area shown in FIG. 11 (S41).

FIG. 11 is a diagram illustrating a configuration of an edge influence degree calculation area set for a certain pixel (target pixel).
Here, in the third embodiment, the filter processing area is composed of 11 × 11 pixels, but filter processing areas of various sizes such as 21 × 21 pixels may be set according to the size conversion magnification of the decoded image.
First, the edge strength calculation unit 31 calculates the edge strength of each peripheral pixel using a Sobel filter or the like (S42).

FIG. 12 is a diagram showing a configuration of a Sobel filter. In particular, FIG. 12A is a diagram showing a configuration of a Sobel filter applied in a horizontal direction, and FIG. 12B is a diagram applied to a vertical direction. FIG. 2 is a diagram showing a configuration of a Sobel filter.
The Sobel filter multiplies the nine pixel values of the upper, lower, left, and right centered on a certain pixel of interest (here, peripheral pixels) by the coefficients of the Sobel filter shown in FIG. 12 and sums up the multiplication results. . Assuming that the total value in the horizontal direction is gHS and the total value in the vertical direction is gVS, the edge strength g of the pixel of interest is obtained by the following equation.
g = (gHS ² + gVS ² ) ^1/2
By performing such processing on each peripheral pixel, it is possible to detect the edge strength of each peripheral pixel.

  In the third embodiment, the Sobel filter is used as the edge strength calculation unit 31. However, the edge strength may be obtained by using a primary differential filter such as a Prewitt filter or a secondary differential filter. Good.

When the calculation of the edge strength of the peripheral pixels is completed, the maximum edge influence degree calculation unit 32 executes a maximum edge influence degree calculation process for obtaining a distribution coefficient to be applied to the filter processing area (S43).
FIG. 13 is a flowchart illustrating a subroutine of a maximum edge influence degree calculation process.
The maximum edge influence degree calculation unit 32 repeatedly executes the following processing for each peripheral pixel in the edge influence degree calculation area (S431).

  First, the maximum edge influence degree calculation unit 32 calculates the distance from each pixel of interest in the edge influence degree calculation area to the target pixel (S432). The distance obtained in this manner is indicated by a numeral in FIG. In the third embodiment, the diagonal distance between the peripheral pixel and the target pixel is simply obtained by adding the horizontal distance and the vertical distance, but the square of the horizontal distance is used. And the square root of the sum of the square of the vertical distance and the square of the vertical distance.

  When the distance is calculated, the maximum edge influence degree calculation unit 32 calculates the edge influence degree based on the distance from the target pixel calculated in step S432 and the edge strength of the peripheral pixel calculated by the edge strength calculation unit 31. (S433). Here, as shown in FIG. 14, the edge influence degree is represented by a function F (distance, edge strength) based on the distance from the pixel of interest and the edge strength. That is, the farther from the pixel of interest and the smaller the edge strength, the smaller the value.

When the edge influence of each peripheral pixel in the filter processing area is obtained by repeatedly executing such processing (S434), the maximum edge influence calculation unit 32 calculates the distribution coefficient to be applied to the filter processing area as a distribution coefficient. The maximum value of the edge influence degree is adopted (S435), and the process returns to the main routine shown in FIG.
When the calculation of the distribution coefficient applied to each filter processing area is completed (S44), the difference value calculation unit 11 repeatedly executes the following processing for each pixel of the decoded image (S16). The difference value calculation unit 11 calculates the pixel value f (x, y) of a certain target pixel and the pixel value f (x, y) of each peripheral pixel in the filter processing area (all the peripheral pixels included in the filter processing area shown in FIG. 4). x + i, y + j) and the difference values ν _{x, y} (i, j) are calculated (S17).

When the calculation of all the difference values v _{x, y} (i, j) is completed (S18), the filter coefficient calculation unit 33 repeatedly executes the following processing for each peripheral pixel in the filter processing area (S19). The filter coefficient calculator 33 calculates the difference value ν _{x, y} (i, j) calculated by the difference value calculator 11 and the distribution coefficient U (X, Y) calculated by the maximum edge influence degree calculator 32. Then, a filter coefficient β _{x, y} (i, j) for each peripheral pixel is calculated (S20). Specifically, the filter coefficient β _x ,, is calculated using an expression in which the distribution coefficient U (X, Y) in Expression (8) is replaced by the distribution coefficient U (X, Y) calculated in the maximum edge influence degree calculation unit 32 _. Calculate _y (i, j).

When all the filter coefficients β _{x, y} (i, j) have been calculated (S21), the filter processing unit 34 repeatedly executes the following processing for each pixel of interest (S22). The filter processing unit 34 performs filter processing on the pixel of interest using the filter coefficient β _{x, y} (i, j) for each peripheral pixel calculated by the filter coefficient calculation unit 33, and performs filtering on the pixel of interest after the filter processing. A value is calculated (S23). When such processing (S23) is repeatedly executed and the pixel values after the filter processing are calculated for each pixel in all the blocks (S24), a restored image is generated, and the restored image generation processing ends.

  As described above, according to the image processing apparatus 3 according to the third embodiment, the edge strength calculation unit 31 calculates the edge strength of the peripheral pixels in the edge influence calculation area, and the maximum edge influence calculation unit 32 A distribution coefficient (edge influence area maximum value) to be applied to the filter processing area is determined according to the distance from the edge pixel and the magnitude of the edge strength, and the filter coefficient calculation unit 33 determines a filter coefficient according to the distribution coefficient. Is calculated, and the filter processing unit 34 changes the smoothing strength of the pixel.

  Therefore, even when the block boundary position is unknown, it can be effectively processed. Moreover, even when a natural image and an artificial image are mixed, blurring of the texture as in the past is prevented, and regardless of whether the edge strength is high or low, blurring of the edge is suppressed, Mosquito noise can be reliably removed. Also, since the filtering process is performed independently of the block boundaries, it is necessary to surely prevent a situation in which the filtered image and the unprocessed image are extremely separated from each other as in the related art, and the image quality suddenly changes. Can be.

  In the image processing apparatuses 1 and 2 of the first and second embodiments, the distribution coefficient U (X, Y) applied to the block is calculated based on the maximum pixel value max {f (x, y)} in the block and the minimum pixel Although the value obtained by subtracting the value min {f (x, y)} is not limited to this value, the value may be any value that indicates the magnitude of the distribution of pixel values in the block. That is, for example, a variance value from an average pixel value in a block can be used as the distribution coefficient U (X, Y). Alternatively, the edge intensity may be calculated for each pixel using a Sobel filter, and the maximum edge intensity in each block may be used as the distribution coefficient U (X, Y).

A method of calculating a filter coefficient β _{x, y} (i, j) based on the difference value ν _{x, y} (i, j) and the distribution coefficient U (X, Y) (or the interpolation distribution coefficient u (x, y)) Is not calculated based on a continuous function as in Equation (6), but instead of the difference value ν _{x, y} (i, j) and the distribution coefficient U (X, Y) (or the interpolation distribution coefficient u ( A plurality of levels may be set in advance for the values of (x, y)), and the values may be set stepwise according to each level.

When a color image is processed, a filter coefficient β _{x, y} (i, j) is calculated from the luminance component (Y), and R, R is calculated using the filter coefficient β _{x, y} (i, j). Filter processing may be performed for each of the G and B components.
Further, the filter processing in the present invention is not limited to the processing represented by Expression (5), and may be realized by another expression as long as the same means is provided.

Further, the present invention is not limited to the process of removing noise generated by JPEG, but can be used for the process of removing noise generated by another block coding method.
Further, the present invention is not limited to still images, and can be applied to moving images as in the case of processing on MPEG compressed data. The same effects as in the case of still images can be obtained in moving images.
Further, the processing need not necessarily be performed inside the printer or the projector, but may be performed on a PC.

  In the first and second embodiments, direct printing is performed by a printer. However, even when printing is performed from a PC, the data is not decoded into RGB data in the PC and then sent to the printer. A method of transferring all data from the PC to the printer and executing all the processing after the decryption processing inside the printer, that is, so-called pull printing, has recently been studied, and can be applied to such a method. Also in this case, since the printer receives the data in the form of JPEG compressed data and performs all the processing internally from the decoding processing, the block boundary position can be naturally known.

  The image processing apparatus includes a filter for removing noise from a decoded image obtained by decoding compressed data obtained by encoding a multi-valued image in units of M × N pixels, such as a printer or a projector. The present invention can be applied to a computer device that performs processing and generates a restored image.

FIG. 2 is a diagram illustrating an external configuration of a digital camera and a printer. FIG. 2 is a block diagram illustrating a functional configuration of an image processing apparatus mounted on the printer ex400 illustrated in FIG. 1. 5 is a flowchart illustrating an operation when the image processing apparatus 1 generates a restored image from a decoded image. FIG. 3 is a diagram illustrating a positional relationship between a target pixel, a filter processing area, and a block. FIG. 4 is a block diagram illustrating a functional configuration of another image processing apparatus mounted on the printer ex400 illustrated in FIG. 1. 6 is a flowchart showing an operation when the image processing apparatus 2 generates a restored image from a decoded image. It is a figure for explaining interpolation processing of a distribution coefficient. FIG. 2 is a diagram illustrating an external configuration of a digital camera, a personal computer, a magnification converter, and a projector. FIG. 9 is a block diagram illustrating a functional configuration of an image processing device mounted on the projector ex600 illustrated in FIG. 8. 5 is a flowchart illustrating an operation of the image processing apparatus 3 when generating a restored image from a decoded image. FIG. 3 is a diagram illustrating a configuration of a filter processing area and an edge influence degree calculation area set for a certain pixel (pixel of interest). It is a figure which shows the structure of a Sobel filter, especially FIG.12 (a) is a figure which shows the structure of the Sobel filter applied in a horizontal direction, FIG.12 (b) is the Sobel applied in a vertical direction. It is a figure showing composition of a filter. It is a flowchart which shows the subroutine of the maximum edge influence degree calculation process. FIG. 9 is a diagram illustrating a relationship among a distance from a target pixel, edge strength, and an edge influence degree. FIG. 3 is a block diagram illustrating a functional configuration of a JPEG encoding device and a decoding device. FIG. 2 is a diagram illustrating a conventional image processing apparatus (SUSAN filter) 100. FIG. 7 is a diagram illustrating a relationship between a difference value between a pixel value of a target pixel and a pixel value of each peripheral pixel, and a value of a second filter coefficient α2 _{x, y} (i, j) calculated from the difference value. Diagram showing gradation of edge and mosquito noise It is a figure which shows the conventional filter processing typically. It is a figure which shows typically the decoded image before performing the filter processing by a SUSAN filter, and the restored image after filter processing.

Explanation of reference numerals

1, 2, 3 image processing device 11 difference value calculation unit 12 distribution coefficient calculation unit 13, 23, 33 filter coefficient calculation unit 14, 24, 34 filter processing unit 22 distribution coefficient interpolation unit 31 edge intensity calculation unit 32 maximum edge influence degree Calculator

Claims (13)

Image processing for generating a restored image by performing filter processing for removing noise on a decoded image obtained by decoding compressed data obtained by encoding a multi-valued image for each block of M × N pixels A device,
A difference value calculation unit that calculates a difference value between a pixel value of a target pixel and a pixel value of a peripheral pixel in a filter processing region set for each pixel configuring the decoded image,
Distribution coefficient calculating means for calculating a distribution coefficient of a pixel value in each of the pixels;
A filter coefficient calculating unit that calculates a filter coefficient for the peripheral pixel based on the difference value calculated by the difference value calculating unit and the distribution coefficient calculated by the distribution coefficient calculating unit;
Filter processing means for performing filter processing on the pixel value of the pixel of interest in the decoded image using the filter coefficient calculated by the filter coefficient calculation means, and calculating the pixel value of the pixel of interest in the restored image. An image processing apparatus comprising: The image processing device according to claim 1, wherein the distribution coefficient calculation unit calculates a distribution coefficient of a pixel value of each of the pixels for each of the blocks. The image processing apparatus according to claim 2, wherein the distribution coefficient calculation unit calculates a difference value obtained by subtracting a minimum value from a maximum value of pixel values in the block as a distribution coefficient of pixel values of the block. . The image processing apparatus according to claim 2, wherein the distribution coefficient calculation unit calculates a variance value from an average pixel value in the block as a distribution coefficient of a pixel value of the block. The image processing device according to claim 2, wherein the distribution coefficient calculation unit calculates a maximum edge intensity in the block as a distribution coefficient of a pixel value of the block. The distribution coefficient calculation unit calculates a distribution coefficient of a pixel value for each block, and interpolates a distribution coefficient of a pixel value in the calculated block and a distribution coefficient of a pixel value in a neighboring block adjacent to the block. The image processing apparatus according to claim 1, wherein an interpolation distribution coefficient of a pixel value of each pixel is calculated using the calculated interpolation distribution coefficient, and the calculated interpolation distribution coefficient is used as a distribution coefficient of a pixel value of each pixel. The image processing apparatus according to claim 6, wherein the distribution coefficient calculation unit calculates a difference value obtained by subtracting a minimum value from a maximum value of pixel values in the block as a distribution coefficient of pixel values of the block. The image processing apparatus according to claim 6, wherein the distribution coefficient calculation unit calculates a variance value from an average pixel value in the block as a distribution coefficient of a pixel value of the block. The image processing apparatus according to claim 6, wherein the distribution coefficient calculation unit calculates a maximum edge intensity in the block as a distribution coefficient of a pixel value of the block. The distribution coefficient calculating means,
Edge strength calculation means for calculating the edge strength of the peripheral pixel in a predetermined edge influence calculation area,
The edge intensity of the peripheral pixel calculated by the edge intensity calculation means, the edge influence degree for the pixel of interest is calculated based on the distance between the pixel of interest and the peripheral pixel, the maximum value of the calculated edge influence degree each The image processing apparatus according to claim 1, further comprising: a maximum edge influence degree calculating unit that calculates a distribution coefficient of a pixel value of the pixel. Image processing for generating a restored image by performing filter processing for removing noise on a decoded image obtained by decoding compressed data obtained by encoding a multi-valued image for each block of M × N pixels An image processing method used for the device,
A difference value calculating step of calculating a difference value between a pixel value of a target pixel and a pixel value of a peripheral pixel in a filter processing region set for each of the pixels constituting the decoded image,
A distribution coefficient calculating step of calculating a distribution coefficient of a pixel value in each of the pixels;
A filter coefficient calculating step of calculating a filter coefficient for the peripheral pixel based on the difference value calculated by the difference value calculating step and the distribution coefficient calculated by the distribution coefficient calculating step;
Performing a filtering process on the pixel value of the target pixel in the decoded image using the filter coefficient calculated in the filter coefficient calculating step, and calculating a pixel value of the target pixel in the restored image. An image processing method comprising:   A program for causing a computer to execute all steps included in the image processing method according to claim 11.   A computer-readable recording medium on which a program for executing all steps of the image processing method according to claim 11 is recorded.

Images