JP2012239033A - Image processing device, decoding device, and image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique for preventing a problem such as deterioration due to a defect of an image or the like due to truncation with a simple circuit even when encoding processing incapable of truncation is used for compression processing of gradation data.SOLUTION: When the code amount of encoded data obtained through packing for a pixel block of interest exceeds a threshold, an irreversible encoding result of a difference image is truncated from the encoded data of the pixel block of interest so that the code amount falls below the threshold. Then, encoded data obtained through the truncation is output as the encoded data of the pixel block of interest.

Description

  The present invention relates to an encoding technique.

  As one of methods for compressing an image, block coding is known. In block coding such as JPEG, since the code amount for each block is variable length, it is not possible to meet the desire to decode only a certain block for editing, and decoding must be performed in order from the top.

  On the other hand, Patent Document 1 discloses an encoding technique in which the code amount is fixed length in units of blocks. Since it has a fixed length, if the position of the target block is known, the storage position of the encoded data can be specified by calculation, and only the target block can be decoded. The encoding technique described in Patent Document 1 separates an image in a block into a portion where sharpness of edges such as characters and line drawings is important and a portion where importance of gradation such as natural images is important. It is compressed with the encoding method. For example, in encoding of characters and line drawings, run length encoding and Huffman encoding are applied in combination as encoding of binary block data (hereinafter referred to as identification information). In natural image encoding, JPEG encoding is applied as encoding multi-value block data (hereinafter referred to as gradation data). Then, fixed-length encoded data for one block is generated according to a predetermined format for the obtained encoded data.

  Another method for image compression is JPEG2000. As a JPEG2000 encoding procedure, first, discrete wavelet transform processing is performed for each encoding unit, and transform coefficients are output. Next, the transform coefficient is quantized by a predetermined quantization step to obtain a quantization index. Next, variable length coding is performed on the quantization index. First, the quantization index is decomposed for each bit plane, and binary arithmetic coding is performed for each bit plane. First, each bit of the most significant bit plane is arithmetically encoded and output as a bit stream. Next, arithmetic coding is performed with the bit plane lowered by one level. Thereafter, similarly, encoding is performed up to the least significant bit plane, and a bit stream is output. In addition, since the code amount can be controlled by appropriately terminating the encoding according to the code amount in the process of variable length encoding, it can be said that fixed-length encoding can be substantially performed.

JP 2008-278042 A

  When a variable-length code such as JPEG is stored in a fixed-length code as in the conventional example, it cannot be guaranteed that it falls within the target code amount, and the target code amount may be exceeded. One of the countermeasures in that case is a method of cutting off codes that exceed the target code amount. When the code is cut off to have a fixed length, the grayscale data stored at the end of the code is cut off. When the gradation data stored at the end of the fixed-length code is cut off, in normal JPEG data, the rightmost MCU (Minimum Coded Unit) in the fixed-length coding unit block Furthermore, it is cut off in order from the high frequency component. When the DC component of the rightmost MCU is censored, the MCU located to the left becomes the next censoring target. At this time, since the code of the lower right MCU is lost, the decoding process cannot be performed in the decoder, and a partial loss appears in the image.

  As another countermeasure, there is a method of using an encoding that can be aborted, such as JPEG2000. However, JPEG2000 is divided by a spatial frequency, the encoding algorithm is complicated, the calculation amount is large, and the circuit after mounting The scale will also increase.

  The present invention has been made in view of such a problem, and even when a non-censored encoding process is used for the gradation data compression process, an image loss due to censoring can be achieved with a simple circuit. It aims at providing the technique for preventing problems, such as degradation by such as.

  In order to achieve the object of the present invention, for example, an image processing apparatus of the present invention comprises the following arrangement. That is, a dividing unit that divides an input image into pixel blocks each composed of a plurality of pixels, and a high-frequency pixel in which each pixel constituting the pixel block has a pixel value equal to or greater than a threshold value for each pixel block divided by the dividing unit. , A first generation unit that generates image information indicating which of the low-frequency pixels has a pixel value smaller than the threshold, and the image information generated by the first generation unit for each pixel block A lossless encoding means for performing lossless encoding, a calculation means for obtaining an average pixel value of pixel values of the high-frequency pixels in the pixel block for each pixel block divided by the dividing means, and the dividing means For each pixel block, a gradation obtained by replacing the pixel value of the high-frequency pixel in the pixel block with a pixel value calculated using the low-frequency pixel in the pixel block A second generating unit that generates an elementary block; an irreversible encoding unit that performs irreversible encoding on a gradation pixel block generated for each pixel block by the second generating unit; and the dividing unit For each pixel block, the average pixel value calculated by the calculating unit, the lossless encoding result by the lossless encoding unit, and the lossy encoding result by the lossy encoding unit are packed, and the pixel block is encoded. Output means for outputting as data, wherein the lossy encoding means generates a reduced image of the gradation pixel block as a representative image from the gradation pixel block, and the representative from the gradation pixel block. A third generation unit configured to generate a difference image obtained by subtracting the representative image from a plurality of reduced images obtained by a method different from the image; and the representative image and the third generation unit Means for generating lossy encoding results for the gradation pixel block by performing lossy encoding on each of the difference images, and the output means performs the pack on the pixel block of interest. If the code amount of the encoded data obtained in this way exceeds a threshold value, the lossy encoding result of the difference image is truncated from the encoded data of the pixel block of interest so that the code amount is below the threshold value, The encoded data obtained by the truncation is output as the encoded data of the pixel block of interest.

  According to the configuration of the present invention, even when encoding processing that cannot be censored is used for compression processing of gradation data, problems such as degradation due to image loss due to censoring can be prevented with a simple circuit. can do.

1 is a block diagram illustrating an example of a functional configuration of an image processing apparatus according to a first embodiment. FIG. 5 is a block diagram showing an example of a functional configuration of a second encoding unit 104. The block diagram which shows the structural example of a subblock encoding part. The block diagram which shows the function structural example of the decoding apparatus which concerns on 1st Embodiment. The block diagram which shows the function structural example of the 2nd decoding part 403. FIG. FIG. 5 is a block diagram illustrating an example of a functional configuration of an image processing apparatus according to a second embodiment. The block diagram which shows the function structural example of the 2nd encoding part 604. FIG. The block diagram which shows the function structural example of the decoding apparatus which concerns on 2nd Embodiment. The block diagram which shows the function structural example of the 2nd decoding part 803. FIG. The figure which shows an example of image information.

  Preferred embodiments of the present invention will be described below with reference to the accompanying drawings. The embodiment described below shows an example when the present invention is specifically implemented, and is one of the specific examples of the configurations described in the claims.

[First Embodiment]
In this embodiment, in order to simplify the description, the image to be encoded has only a density component, and the pixel value (density value) of each pixel is represented by 8 bits (0 to 255). To do. That is, in this embodiment, the image to be encoded is a monochrome grayscale image. Note that since the pixel value of each pixel indicates a density value, the larger the value is, the blacker the color is, and vice versa. In addition, when this embodiment is applied to a color image, it is obvious that the following processing is performed for each color component. A functional configuration example of the image processing apparatus according to the present embodiment will be described with reference to the block diagram of FIG.

  An image to be encoded is input to the blocking unit 101 as an input image. The blocking unit 101 divides the input image into pixel blocks each having a size of horizontal m pixels × vertical n pixels (that is, a plurality of pixels). Then, the blocking unit 101 sequentially sends the divided pixel blocks to the extraction unit 102.

  For each pixel block received from the blocking unit 101, the extraction unit 102 includes, for each pixel constituting the pixel block, a high-frequency pixel having a pixel value greater than or equal to a threshold value and a low-frequency pixel having a pixel value smaller than the threshold value. Image information indicating which one is generated is generated (first generation). The high frequency pixel is a pixel that causes high frequency in the pixel block, and examples thereof include pixels constituting characters, line drawings, and the like. When generating image information for the pixel block of interest, the extraction unit 102 refers to the pixel value of each pixel constituting the pixel block of interest, and if the pixel value of the pixel is greater than or equal to the threshold value, 1 "is output. On the other hand, when the pixel value of the pixel is smaller than the threshold value, “0” is output as the bit value for the pixel. In this way, a set of bit values output for each pixel constituting the target pixel block is image information of the target pixel block. Since the bit value for each pixel only needs to be able to indicate whether each pixel is a high frequency pixel or a low frequency pixel, the number of bits is one. In this case, the image information has a size of m bits × n bits. Of course, the bit value is not limited to such a value. Then, the extraction unit 102 sends the image information generated for each pixel block to the first encoding unit 103.

  For each pixel block received from the blocking unit 101, the extraction unit 102 obtains an average value (average pixel value) of high-frequency pixels in the pixel block (calculation). Then, the extraction unit 102 sends the average pixel value obtained for each pixel block to the pack unit 105 as block header data. If no high-frequency pixel exists in the pixel block, the extraction unit 102 sends an appropriate predetermined value to the pack unit 105 as an average pixel value. The reason why it is an appropriate value determined in advance is that the average pixel value is not used when there is no high-frequency pixel in the pixel block.

  In addition, for each pixel block received from the blocking unit 101, the extraction unit 102 replaces the pixel value of the high-frequency pixel in the pixel block with a pixel value calculated using the low-frequency pixel in the pixel block. Are generated as gradation pixel blocks (second generation). For example, a pixel block in which the pixel value of the high frequency pixel in the pixel block is replaced with the average pixel value of the low frequency pixel in the pixel block is generated as a gradation pixel block. The gradation pixel block generated in this way is a pixel block having a size of m pixels × n pixels substantially having no high frequency portion. Then, the extraction unit 102 sends the gradation pixel block generated for each pixel block to the second encoding unit 104.

  When receiving the image information sent for each pixel block from the extraction unit 102, the first encoding unit 103 performs lossless encoding on the image information to generate encoded data, and the generated encoded data The code amount (number of bits) is sent to the pack unit 105. The type of lossless encoding performed by the first encoding unit 103 is not particularly limited, but in this embodiment, run-length encoding is used as an example.

  When the second encoding unit 104 receives the gradation pixel block transmitted from the extraction unit 102 for each pixel block, the second encoding unit 104 performs lossy encoding on the gradation pixel block to generate encoded data, The generated encoded data is sent to the pack unit 105.

  The pack unit 105 packs the block header data transmitted from the extraction unit 102 and the encoded data transmitted from the first encoding unit 103 and the second encoding unit 104 in accordance with an algorithm described later. Output as digitized data.

  The above is a rough process performed by each unit constituting the image processing apparatus according to the present embodiment. Hereinafter, the blocking unit 101, the extraction unit 102, the second encoding unit 104, and the pack unit 105 will be described in more detail.

  The blocking unit 101 has a buffer memory for n lines inside. The input images are input to the block forming unit 101 in the raster scan order and stored in this buffer memory. Then, when n lines of input images are stored in the buffer memory, the blocking unit 101 sequentially cuts out images (pixel blocks) of m pixels in the horizontal direction and n pixels in the vertical direction and outputs them to the extraction unit 102. repeat. In the present embodiment, it is assumed that m = n = 16, that is, the pixel block size is 16 pixels in the horizontal direction and 16 pixels in the vertical direction. However, in the following description, m and n may be any integer values (of course, the input image size or less), and m and n may be different values. Further, m and n may be set as appropriate by the user.

  When the extraction unit 102 receives a pixel block having a size of 16 pixels × 16 pixels from the blocking unit 101, the extraction unit 102 refers to the pixel value of each pixel constituting the pixel block. Each pixel constituting the pixel block is either a high frequency pixel or a low frequency pixel, such that a pixel having a pixel value equal to or higher than the threshold TH is a high frequency pixel, a pixel having a pixel value smaller than the threshold TH is a low frequency pixel, and so on. Classify. The threshold value TH may be obtained by any method as long as a high-frequency pixel can be specified. That is, the user may set in advance, or the extraction unit 102 may obtain from the input image. When it is determined that a pixel having a low density value is a high frequency pixel, a pixel value equal to or lower than the threshold value TH may be set as a high frequency pixel, and a pixel value higher than the threshold value TH may be set as a low frequency pixel. As an example of a method for obtaining the threshold value TH from the input image, there is a method in which an average value of pixel values of each pixel constituting the input image is set as the threshold value TH. Then, the extraction unit 102 sends the average value (average pixel value) AVE1 of the pixel values of the high frequency pixels to the pack unit 105. As described above, when no high-frequency pixel exists in the pixel block, the extraction unit 102 sends a predetermined appropriate value to the pack unit 105.

  Further, when the extraction unit 102 receives a pixel block having a size of 16 pixels × 16 pixels from the blocking unit 101, the extraction unit 102 scans the pixel block and refers to the pixel value of each pixel. The extraction unit 102 outputs a bit value “1” if the referenced pixel value is equal to or greater than the threshold value TH, and outputs a bit value “0” if the pixel value is smaller than the threshold value TH. In this manner, the bit values output for each pixel constituting the pixel block are the number of pixels (m × n), and the extraction unit 102 performs first encoding using a set of bit values for the number of pixels as image information. The data is sent to the unit 103.

  In addition, when the extraction unit 102 receives a pixel block having a size of 16 pixels × 16 pixels from the blocking unit 101, the extraction unit 102 calculates an average value (average pixel value) AVE2 of low-frequency pixels in the pixel block. Then, the extraction unit 102 generates a pixel block in which the pixel value of the high frequency pixel in the pixel block is replaced with the average pixel value AVE2 as a gradation pixel block.

  The pixel value of the pixel at the coordinates (i, j) in the pixel block received by the extraction unit 102 is Ain (i, j), the bit value output by the extraction unit 102 for this pixel is P (i, j), the gradation Let Aout (i, j) be the pixel value of the pixel at coordinates (i, j) in the pixel block. At this time, Aout (i, j) is determined according to the following equation.

P (i, j) = 1 (when the pixel at coordinates (i, j) in the pixel block is a high-frequency pixel) Aout (i, j) = AVE2
P (i, j) = 0 (when the pixel at coordinates (i, j) in the pixel block is a low frequency pixel) Aout (i, j) = Ain (i, j)
Then, the extraction unit 102 sends the gradation pixel block generated in this way to the second encoding unit 104. As a result, the second encoding unit 104 is supplied with the gradation pixel block from which the high-frequency component is excluded, and a high compression rate can be expected. Note that the gradation pixel block may be generated by the second encoding unit 104.

  A functional configuration example of the second encoding unit 104 will be described with reference to the block diagram of FIG. Gradation pixel blocks are sequentially input from the extraction unit 102 to the division unit 201. The dividing unit 201 generates a reduced image of the gradation pixel block from the input gradation pixel block as a representative image, and a plurality of reduced images obtained from the gradation pixel block by a method different from the representative image. A difference image is generated by subtracting the representative image from (a third generation). In the case of FIG. 2, a total of four sub-blocks, one representative image and three difference images, are generated from one gradation pixel block. Of course, the vertical and horizontal sizes of the representative image and the difference image are the same. The dividing unit 201 inputs the generated representative image to the sub-block encoding unit 202a, and inputs the generated three difference images to the sub-block encoding units 202b to 202d, respectively.

  Here, a method for generating the representative image and the difference image will be described. In the following, a case where four sub-blocks are generated will be described. However, the number of sub-blocks is not limited to four. If one representative image and one or more difference images can be generated, The number of subblocks may be any number. In this case, it is necessary to prepare as many subblock encoders as the number of subblocks. However, the configuration may be such that only one sub-block encoding unit is provided, and this one sub-block encoding unit sequentially processes each sub-block in time division.

  The size of each sub-block generated from the gradation pixel block is suitably the same as the size for orthogonal transform processing performed in the subsequent sub-block encoding unit. Here, in order to perform processing including 8 × 8 DCT as orthogonal transform coding, a sub-block having a size of 8 pixels × 8 pixels is generated from a gradation pixel block having a size of 16 pixels × 16 pixels.

  First, a representative image that is a reduced image of a gradation pixel block is generated, but the method of generating a representative image is not limited to the method described below. However, it is not possible to use anything that cannot be calculated in reverse, such as the median.

  If the pixel value at coordinates (i, j) in the gradation pixel block is A (i, j) and the pixel value at coordinates (i, j) in the representative image is B0 (i, j), then B0 (i, j) j) is obtained according to the following equation. Note that i and j are integers satisfying 0 ≦ i ≦ 7 and 0 ≦ j ≦ 7.

B0 (i, j) = (A (2i, 2j) + A (2i + 1,2j) + A (2i, 2j + 1) + A (2i + 1,2j + 1)) / 4
Next, the pixel value at the coordinate (i, j) in the difference image B1 is B1 (i, j), the pixel value at the coordinate (i, j) in the difference image B2 is B2 (i, j), and the difference image B3. If the pixel value at the middle coordinate (i, j) is B3 (i, j), each is obtained according to the following equation.

B1 (i, j) = A (2i + 1,2j) -B0 (i, j)
B2 (i, j) = A (2i, 2j + 1) -B0 (i, j)
B3 (i, j) = A (2i + 1,2j + 1) -B0 (i, j)
That is, the difference image B1 is a difference image obtained by subtracting the representative image B0 from the reduced image composed of A (2i + 1, 2j) with i and j satisfying 0 ≦ i ≦ 7 and 0 ≦ j ≦ 7. The difference image B2 is a difference image obtained by subtracting the representative image B0 from the reduced image composed of A (2i, 2j + 1) with i and j satisfying 0 ≦ i ≦ 7 and 0 ≦ j ≦ 7. The difference image B3 is a difference image obtained by subtracting the representative image B0 from the reduced image composed of A (2i + 1, 2j + 1) with i and j satisfying 0 ≦ i ≦ 7 and 0 ≦ j ≦ 7. The range of pixel values obtained in this way satisfies the following inequality.

0 ≦ B0 (i, j) ≦ 255
−255 ≦ B1 (i, j) ≦ 255
−255 ≦ B2 (i, j) ≦ 255
−255 ≦ B3 (i, j) ≦ 255
Each of the sub-block encoding units 202a to 202d performs lossy encoding on the input sub-block. Considering truncation as an encoding method, encoding using orthogonal transformation such as JPEG is suitable. A configuration applicable to the sub-block encoding units 202a to 202d is shown in FIG. In FIG. 3, the sub-block coding units 202 a to 202 d are used as the sub-block coding unit 202.

  The DCT calculation unit 301 performs DCT conversion processing on the input sub-block. The quantization unit 302 performs a quantization process on the coefficient obtained by the DCT transform. The zigzag scanning unit 303 performs zigzag scanning on each quantized value obtained by the quantization process, and the entropy encoding unit 304 performs entropy encoding on the scanned quantized value to perform encoding. Generate and output sub-block data. A series of encoding processing performed by the sub-block encoding unit 202 is the same as general JPEG encoding, but differs in the following points. Since the pixel value of the image is normally input as the input of the DCT calculation unit 301, a negative value is never input. However, in the present embodiment, as described above, a negative value may be input to input the representative image B0 and the difference images B1 to B3. Therefore, the DCT calculation unit 301 needs to support input of a negative value.

  The pack unit 105 includes block header data sent from the extraction unit 102, coded data sent from the first coding unit 103, coded data sent from the second coding unit 104 (representative images B0, The encoded data of the difference images B1 to B3) are packed in this order. Here, the pack unit 105 monitors the code amount of the encoded data, which is the result of packing, for each pixel block. When the code amount of the encoded data obtained by packing the pixel block of interest exceeds the threshold value, the pack unit 105 encodes the encoded data of the pixel block of interest so that the code amount falls below the threshold value. The lossy encoding result of the difference image is cut off (cut off). For example, when the code amount is counted sequentially from the beginning of the encoded data of the pixel block of interest, it is assumed that the code amount exceeds the threshold in the middle of the lossy encoding result of the difference image B2. In this case, the difference image B3 (the remaining lossy encoding result of the difference image B2 may be added thereto) from the encoded data of the pixel block of interest. As a result, fixed-length encoded data whose code amount does not exceed the threshold value can be output for each pixel block.

  Next, a case where the encoded data for each pixel block obtained as described above is decoded will be described. A functional configuration example of an image processing apparatus (decoding apparatus) that performs this decoding will be described with reference to the block diagram of FIG. Note that the image processing apparatus in FIG. 4 may be the same apparatus as the image processing apparatus in FIG. 1 or a different apparatus.

  When header analysis unit 401 receives (acquires) the fixed-length encoded data of the pixel block of interest, it extracts block header data from the fixed-length encoded data. When receiving the encoded data of the pixel block of interest, the first decoding unit 402 decodes the image information encoded by the first encoding unit 103. When receiving the encoded data of the pixel block of interest, the second decoding unit 403 decodes the gradation pixel block encoded by the second encoding unit 104. The restoration unit 404 uses the block header data obtained from the header analysis unit 401, the image information decoded by the first decoding unit 402, and the gradation pixel block decoded by the second decoding unit 403, as a target pixel. Decode and output the block.

  Below, each part shown in FIG. 4 is demonstrated in detail. When the header analysis unit 401 receives the fixed-length encoded data of the pixel block of interest, the header analysis unit 401 extracts block header data from the fixed-length encoded data. This indicates a correct value. Which is indicated can be determined by whether or not the image information decoded by the first decoding unit 402 indicates that “the bit values for all the pixels are 0”. That is, when the bit values for all the pixels are 0, it indicates that there is no high frequency pixel, so the block header data indicates an appropriate value. Then, the header analysis unit 401 sends the extracted block header data to the restoration unit 404. Further, the header analysis unit 401 cues the first bit of the encoded data of the image information from the fixed length encoded data and outputs it to the first decoding unit 402. Of course, instead of performing cue processing, a pointer indicating the position of the first bit may be output together with the fixed-length encoded data.

  The first decoding unit 402 receives fixed-length encoded data in which encoded data of image information is cued, or a pointer indicating the head position of encoded data of image information and fixed-length encoded data. The first decoding unit 402 restores the image information by performing a decoding process (first decoding) corresponding to the first encoding unit 103 on the encoded data of the image information. The image information is sent to the restoration unit 404. In addition, the first decoding unit 402 cues the first bit of the encoded data of the gradation pixel block and outputs it to the second decoding unit 403. Here, instead of the cue, a pointer indicating the head bit position of the encoded data of the gradation pixel block and the fixed-length encoded data may be output together.

  The second decoding unit 403 includes fixed-length encoded data in which the encoded data of the gradation pixel block is cued, or a pointer indicating the start position of the encoded data of the gradation pixel block and the fixed-length encoded data Receive. Then, the second decoding unit 403 performs a decoding process (second decoding) corresponding to the second encoding unit 104 on the encoded data of the gradation pixel block, thereby the gradation pixel block. And the restored gradation pixel block is sent to the restoration unit 404.

  A functional configuration example of the second decoding unit 403 is shown in FIG. In FIG. 5, four sub-block decoding units are provided in accordance with the configuration of FIG. Any of the sub-block decoding units 501a to 501d performs the same decoding process, the decoded sub-block to the synthesizing unit 502, and the data that cues the next sub-block to the next-stage sub-block decoding unit 501. Output. Of course, like the encoding side, one subblock decoding unit may decode each subblock in a time division manner.

  The synthesizing unit 502 decodes the gradation pixel block from the sub-blocks output from the respective sub-block decoding units 501a to 501d, and sends the decoded gradation pixel block to the restoration unit 404.

  Next, each of the sub-block decoding units 501a to 501d and the combining unit 502 will be described in more detail. The sub-block decoding unit 501a obtains the representative image C0 by decoding the encoded data of the representative image B0. The sub-block decoding unit 501b obtains the difference image C1 by decoding the encoded data of the difference image B1. The sub-block decoding unit 501c obtains the difference image C2 by decoding the encoded data of the difference image B2. The sub-block decoding unit 501d obtains the difference image C3 by decoding the encoded data of the difference image B3. Any decoding is a decoding process corresponding to the sub-block encoding unit.

  The synthesizing unit 502 generates a gradation pixel block D from the representative image C0 and the difference images C1 to C3 according to the following formula. i and j are integers satisfying 0 ≦ i ≦ 7 and 0 ≦ j ≦ 7.

D (2i, 2j) = C0 (i, j) -C1 (i, j) -C2 (i, j) -C3 (i, j)
D (2i + 1, 2j) = C0 (i, j) + C1 (i, j)
D (2i, 2j + 1) = C0 (i, j) + C2 (i, j)
D (2i + 1,2j + 1) = C0 (i, j) + C3 (i, j)
C (i, j) indicates a pixel value at coordinates (i, j) in the sub-block. By calculating such an expression, the pixel value at each pixel position in the gradation pixel block can be obtained. Note that, for example, when the difference image C3 is censored by the censoring, C3 (i, j) = 0 is assumed, and calculation is performed as follows.

D (2i, 2j) = C0 (i, j) -C1 (i, j) -C2 (i, j)
D (2i + 1, 2j) = C0 (i, j) + C1 (i, j)
D (2i, 2j + 1) = C0 (i, j) + C2 (i, j)
D (2i + 1,2j + 1) = C0 (i, j)
Of course, when the difference images C1 and C2 are cut off, C1 (i, j) = C2 (i, j) = 0 is similarly set. When the difference images C1 to C3 are censored, D (2i, 2j) to D (2i + 1, 2j + 1) are all C0, which is equivalent to a simple four-pixel average image obtained by orthogonal transform coding. It becomes. Also, these indicate that even if the difference image is cut off, the image can be formed even though the image quality is deteriorated, and the loss of the image can be prevented.

  The restoration unit 404 refers to the image information obtained from the first decoding unit 402, and the bit value corresponds to “1” among the pixels in the gradation pixel block obtained from the second decoding unit 403. A pixel to be operated, that is, a high frequency pixel is specified. Then, the restoration unit 404 outputs a gradation pixel block in which the pixel value of the identified high-frequency pixel is replaced with the average pixel value indicated by the block header data obtained from the header analysis unit 401.

  In this way, by dividing and encoding the representative image and the difference image and storing them, even if the data amount exceeds the target code amount and part of the data is truncated, it is possible to prevent image loss It becomes.

[Second Embodiment]
In the present embodiment, an image processing apparatus that performs encoding with less image quality degradation than the first embodiment will be described. A functional configuration example of the image processing apparatus according to the present embodiment will be described with reference to the block diagram of FIG.

  In FIG. 6, the same reference numerals are assigned to the same components as those shown in FIG. The configuration shown in FIG. 6 is different from the configuration shown in FIG. 1 in that the second encoding unit 104 is replaced with the second encoding unit 604 and the image information generated by the extraction unit 102 is the second. This is also supplied to the encoding unit 604.

  A functional configuration example of the second encoding unit 604 will be described with reference to the block diagram of FIG. Hereinafter, only points different from the second encoding unit 104 will be described, and the same points will not be touched.

  The dividing unit 701 receives the gradation pixel block and the image information corresponding to the gradation pixel block from the extraction unit 102. In the gradation pixel block, the pixel value of the pixel that was originally a high frequency pixel is replaced with the average pixel value of the low frequency pixel. However, this average pixel value does not necessarily match the average pixel value in the region of 2 pixels × 2 pixels referred to obtain the representative image B0. However, if the average pixel value of the low-frequency pixels is included in this 2 pixel × 2 pixel region, a difference occurs between the other 2 pixel × 2 pixel regions, and the sub-block in the subsequent stage There is a possibility that the compression efficiency is lowered in the encoding unit. For this reason, in this embodiment, when the average pixel value of the low-frequency pixel is included in the 2 pixel × 2 pixel region referred to obtain the representative image B0, the average pixel value is not used. .

  Specifically, the dividing unit 701 calculates B0 (i, j) according to the following formula. However, when P (2i, 2j) = P (2i + 1, 2j) = P (2i, 2j + 1) = P (2i + 1, 2j + 1) = 1, this equation is not calculated and B0 (i, j) = AVE2 To do.

B0 (i, j) = ((1-P (2i, 2j)) × A (2i, 2j)
+ (1−P (2i + 1,2j)) × A (2i + 1,2j)
+ (1-P (2i, 2j + 1)) × A (2i, 2j + 1)
+ (1-P (2i + 1,2j + 1)) * A (2i + 1,2j + 1)) / (4-P (2i, 2j) -P (2i + 1,2j) -P (2i, 2j + 1) -P (2i + 1,2j + 1) )
That is, only the pixel value of the pixel whose bit value is 0 is used for the calculation for obtaining the pixel value of the representative image B0. An example is shown in FIG. FIG. 10 shows image information having a size of 16 pixels × 16 pixels. In the case of this image information, B0 (0, 0) (the upper left corner is i = j = 0) is obtained according to the following equation.

B0 (0,0) = (A (1,0) + A (1,1)) / 2
Further, B0 (1,0) is obtained according to the following formula.

B0 (1,0) = (A (2,0) + A (2,1) + A (3,1)) / 3
For B0 (2,0), P (4,0) = P (4,1) = P (5,0) = P (5,1) = 1, so B0 (2,0) = AVE2.

  Similar to the first embodiment, the difference images B1 to B3 are obtained according to the following equations.

B1 (i, j) = A (2i + 1,2j) -B0 (i, j)
B2 (i, j) = A (2i, 2j + 1) -B0 (i, j)
B3 (i, j) = A (2i + 1,2j + 1) -B0 (i, j)
Next, a case where the encoded data for each pixel block obtained as described above is decoded will be described. A functional configuration example of the image processing apparatus that performs this decoding will be described with reference to the block diagram of FIG. Note that the image processing apparatus in FIG. 8 may be the same apparatus as the image processing apparatus in FIG. 6 or a different apparatus. In the configuration shown in FIG. 8, the same reference numerals are given to the same components as those shown in FIG. The configuration shown in FIG. 8 is different from the configuration shown in FIG. 4 in that the second decoding unit 403 is replaced with the second decoding unit 803 and the decoding result (image information) by the first decoding unit 402 is changed. ) Is supplied to the second decryption unit 803.

  A functional configuration example of the second decoding unit 803 will be described with reference to the block diagram of FIG. In the configuration shown in FIG. 9, the same reference numerals are assigned to the same components as those shown in FIG. The configuration shown in FIG. 9 is different from the configuration shown in FIG. 5 in that the synthesis unit 502 is replaced with a synthesis unit 902 and decoded image information is supplied to the synthesis unit 902. . Hereinafter, the combining unit 902 will be described.

  The synthesizing unit 902 decodes the gradation pixel block from the sub-blocks output from the respective sub-block decoding units 501a to 501d, and sends the decoded gradation pixel block to the restoration unit 404.

  The synthesizing unit 902 generates a gradation pixel block D from the representative image C0, the difference images C1 to C3, and the image information according to the following formula. i and j are integers satisfying 0 ≦ i ≦ 7 and 0 ≦ j ≦ 7.

D (2i, 2j) = (1-P (2i, 2j)) * C0 (i, j)
− (1−P (2i + 1,2j)) × C1 (i, j)
− (1−P (2i, 2j + 1)) × C2 (i, j)
− (1−P (2i + 1,2j + 1)) × C3 (i, j)
D (2i + 1, 2j) = C0 (i, j) + C1 (i, j)
D (2i, 2j + 1) = C0 (i, j) + C2 (i, j)
D (2i + 1,2j + 1) = C0 (i, j) + C3 (i, j)
When the image information is as shown in FIG. 10, for example, each of D (0,0) and D (2,0) is calculated as follows.

D (0,0) = 0-C1 (0,0) -0-C3 (0,0)
D (2,0) = C0 (1,0) -0-C2 (1,0) -C3 (1,0)
Note that the difference image that has been censored is dealt with by removing it from the calculation formula, as in the first embodiment. For example, when the difference image C3 is truncated, the term C3 is removed from the above calculation formula.

  As described above, according to the present embodiment, since the average image is calculated from only the low-frequency pixels, the influence of the high-frequency pixels on the gradation information is reduced, so that image coding with less deterioration from the original image is performed. Is possible.

[Third Embodiment]
Each part shown in FIGS. 1 to 9 may be configured by hardware, but may be implemented by a computer program. In the latter case, this computer program is installed in this memory of a computer having a CPU, a memory, etc., and this computer executes this computer program, so that this computer has the functions of the components shown in FIGS. Can be realized. Of course, the computer programs on the encoding side and the decoding side may be installed and executed on separate computers.

(Other examples)
The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, or the like) of the system or apparatus reads the program. It is a process to be executed.

Claims (7)

A dividing unit that divides the input image into pixel blocks each including a plurality of pixels;
For each pixel block divided by the dividing means, whether each pixel constituting the pixel block is a high-frequency pixel having a pixel value equal to or higher than a threshold value or a low-frequency pixel having a pixel value smaller than the threshold value. First generating means for generating image information to be shown;
Lossless encoding means for performing lossless encoding on image information generated by the first generation means for each pixel block;
For each pixel block divided by the dividing means, calculating means for obtaining an average pixel value of pixel values of the high-frequency pixels in the pixel block;
A gradation pixel block obtained by replacing the pixel value of the high-frequency pixel in the pixel block with a pixel value calculated using the low-frequency pixel in the pixel block for each pixel block divided by the dividing means Second generating means for generating
Irreversible encoding means for performing irreversible encoding on the gradation pixel block generated by the second generation means for each pixel block;
For each pixel block divided by the dividing means, the average pixel value calculated by the calculating means, the lossless encoding result by the lossless encoding means, and the lossy encoding result by the lossy encoding means are packed, Output means for outputting as pixel block encoded data,
The lossy encoding means includes
Means for generating a reduced image of the gradation pixel block as a representative image from the gradation pixel block;
Third generation means for generating a differential image obtained by subtracting the representative image from a plurality of reduced images obtained by a method different from the representative image from the gradation pixel block;
Means for generating an irreversible encoding result for the gradation pixel block by performing irreversible encoding on the representative image and each difference image generated by the third generation unit;
The output means includes
When the code amount of the encoded data obtained by performing the pack on the pixel block of interest exceeds a threshold value, the difference image is not detected from the encoded data of the pixel block of interest so that the code amount falls below the threshold value. An image processing apparatus, wherein the lossless encoding result is aborted, and the encoded data obtained by the abort is output as the encoded data of the pixel block of interest.   The second generation unit is obtained by replacing the pixel value of the high frequency pixel in the pixel block with the average pixel value of the low frequency pixel in the pixel block for each pixel block divided by the dividing unit. The image processing apparatus according to claim 1, wherein a gradation pixel block to be generated is generated. Means for obtaining encoded data of each pixel block encoded by the image processing apparatus according to claim 1;
Means for extracting an average pixel value of high-frequency pixels calculated by the calculation means for the target pixel block from the encoded data of the target pixel block;
First decoding means for decoding image information generated by the first generation means for the target pixel block from the encoded data of the target pixel block;
Second decoding means for decoding the representative image and the difference image from the encoded data of the pixel block of interest and generating a gradation pixel block of the pixel block of interest from the decoded representative image and difference image When,
The high-frequency pixel in the gradation pixel block decoded by the second decoding unit is specified from the image information decoded by the first decoding unit, and the pixel value of the specified high-frequency pixel is determined by the first decoding unit. And a means for outputting the gradation pixel block replaced with the decoded average pixel value. An image processing method performed by an image processing apparatus,
A dividing step in which the dividing unit of the image processing apparatus divides the input image into pixel blocks each including a plurality of pixels;
For each pixel block divided by the dividing step by the first generation unit of the image processing device, each pixel constituting the pixel block has a high-frequency pixel having a pixel value equal to or higher than a threshold value, and a pixel value smaller than the threshold value A first generation step of generating image information indicating which is a low-frequency pixel having:
A lossless encoding step in which the lossless encoding means of the image processing device performs lossless encoding on the image information generated for each pixel block in the first generation step;
A calculating step of calculating an average pixel value of pixel values of the high-frequency pixels in the pixel block for each pixel block divided in the dividing step by the calculation means of the image processing device;
The second generation means of the image processing device calculates the pixel value of the high frequency pixel in the pixel block for each pixel block divided in the division step using the low frequency pixel in the pixel block. A second generation step of generating a gradation pixel block obtained by replacing with a pixel value;
An irreversible encoding step in which the irreversible encoding means of the image processing device performs irreversible encoding on the gradation pixel block generated for each pixel block in the second generation step;
The output means of the image processing device, for each pixel block divided in the division step, the average pixel value calculated in the calculation step, the lossless encoding result in the lossless encoding step, the lossy encoding step An irreversible encoding result is packed and output as encoded data of the pixel block, and
The lossy encoding step includes:
Generating a reduced image of the gradation pixel block from the gradation pixel block as a representative image;
A third generation step of generating a differential image obtained by subtracting the representative image from a plurality of reduced images obtained by a method different from the representative image from the gradation pixel block;
Generating lossy encoding results for the gradation pixel block by performing lossy encoding on the representative image and each difference image generated in the third generation step,
In the output step,
When the code amount of the encoded data obtained by performing the pack on the pixel block of interest exceeds a threshold value, the difference image is not detected from the encoded data of the pixel block of interest so that the code amount falls below the threshold value. An image processing method comprising: truncating a lossless encoding result and outputting encoded data obtained by the truncation as encoded data of the pixel block of interest. An image processing method performed by a decoding device,
The acquisition unit of the decoding apparatus acquires encoded data of each pixel block encoded by the image processing apparatus according to claim 1;
Means for extracting, from the encoded data of the target pixel block, an average pixel value of the high-frequency pixels calculated by the calculation means for the target pixel block;
A first decoding step in which a first decoding unit of the decoding device decodes image information generated in the first generation step for the target pixel block from encoded data of the target pixel block;
Second decoding means of the decoding device decodes the representative image and the difference image from the encoded data of the pixel block of interest, and calculates the floor of the pixel block of interest from the decoded representative image and difference image. A second decoding step for generating a tonal pixel block;
The output unit of the decoding device specifies the high frequency pixel in the gradation pixel block decoded in the second decoding step from the image information decoded in the first decoding step, and the pixel value of the specified high frequency pixel And a step of outputting the gradation pixel block in which the average pixel value decoded in the first decoding step is replaced.   A computer program for causing a computer to function as each unit of the image processing apparatus according to claim 1.   The computer program for functioning a computer as each means of the decoding apparatus of Claim 3.

Images