JP3781012B2 - Image data compression method, image data expansion method, and image data expansion circuit - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to image data compression / decompression technology using wavelet transform, and particularly to an image data compression method, image data expansion method, and image data expansion circuit that enable high-speed expansion processing.
[0002]
[Prior art]
As an image compression technique standardized by JPEG (Joint Photographic Picture Experts Group), a technique combining DCT transform (discrete cosine transform) and Huffman coding is known. In this image compression technique, the accuracy of quantization of image data is coarsened in a frequency band that is insensitive to vision, and a short code is assigned to image data having a high appearance frequency, thereby reducing the total amount of data. (Refer nonpatent literature 1).
JPEG2000 employs an image compression technique that combines wavelet transform and arithmetic coding. According to this technique, a higher compression ratio is achieved compared to the image compression technique standardized by JPEG described above. However, image degradation can be suppressed to a small level (see Non-Patent Document 2).
[0003]
[Non-Patent Document 1]
“Latest MPEG Textbook”, ASCII Publishing, first published August 1, 1994, p. 53-67
[Non-Patent Document 2]
bit-drive [Broadband encyclopedia], [March 10, 2003 search], Internet <URL: http: //bit-drive.e-words.ne.jp/w/JPEG2000.html>
[0004]
[Problems to be solved by the invention]
By the way, according to the above-described image compression technique standardized by JPEG, DCT conversion is performed in units of 8 × 8 pixel size blocks. Therefore, when the compression rate is increased, distortion becomes conspicuous in units of blocks. Moreover, according to the DCT transform, an image is approximated using a sine wave, and therefore, when an edge is strong and a spatial frequency is high as in an animation image, the approximation accuracy is lowered, and so-called mosquito noise may occur. . In addition, according to Huffman coding, a table for associating the image data with the coded code is required. In general, a method of referring to a table has a high affinity with software and is not suitable for hardware implementation, and therefore has a problem that each process of encoding (compression) and decoding (decompression) takes time.
[0005]
In addition, according to the image compression technique standardized by the above-described JPEG2000, in order to perform wavelet transform on a two-dimensional plane, a memory having a large capacity equivalent to the pixel size of the screen must be prepared as a buffer. Therefore, it is not suitable for hardware implementation, and the processing on software is the main, so that there is a problem that it takes time for the decryption processing. Moreover, arithmetic coding requires complicated arithmetic processing, and particularly when encoding is performed in units of bit planes as in JPEG2000, there is a problem that it takes a long time for decoding processing.
[0006]
The present invention has been made in view of the above circumstances, and enables image data compression method and image data decompression that enable image data to be decompressed at a high speed while suppressing the influence on the image quality, and that facilitate hardware implementation. It is an object to provide a method and an image data decompression circuit.
[0007]
[Means for Solving the Problems]
  In order to solve the above problems, the present invention has the following configuration.
  According to the present inventionIn the image data compression method, the image data to be compressed is divided into a plurality of blocks, the image data is subjected to wavelet transform for each block, and the image data of each block subjected to the wavelet transform is image data indicating a zero value. Each block in accordance with a predetermined format that defines a correspondence relationship when converting the first data of 0 run length indicating how many consecutive to the second data representing the value of the image data itself indicating the non-zero value. Replacing the image data with the first or second data, and encoding.
[0008]
According to this configuration, the block-divided image data is subjected to wavelet transform and separated into a visually sensitive low frequency component and an insensitive high frequency component. The image data from which the frequency band has been separated is encoded into 0 run-length data indicating how many consecutive image data indicating 0 values, or data indicating the value of image data indicating non-zero values. Replaced. Here, according to the wavelet transform, there is a property that the value of the image data of the high frequency component becomes small, and the appearance frequency of “0” as the value of the image data tends to increase. For this reason, the frequency of replacement with data representing 0 run length at the time of encoding increases, thereby compressing the data. Therefore, according to this configuration, it is sufficient to process the data representing the 0 run length or bit string of the image data when compressing, and it is not necessary to perform complicated operations. In this case, it is not necessary to perform a complicated operation in the same manner, so that high-speed expansion is possible.
[0009]
  The image data compression method further includes the step of converting the image data from an RGB system to a YUV system before the wavelet transform.
  The image data compression method further includes the step of calculating the difference of the image data between adjacent pixels after quantizing the wavelet-transformed image data.
  That is, the image data compression method according to the present invention includes a first step of dividing image data to be compressed into a plurality of blocks, a second step of converting the divided image data from an RGB system to a YUV system, A wavelet transform of the image data converted into the YUV system for each block, a third step of thinning out the color difference component at that time, and a fourth step of quantizing the transform coefficient of the image data obtained by the wavelet transform; A fifth step of calculating the difference between the quantized transform coefficients and the transform coefficients between adjacent pixels, and image data including the difference between the transform coefficients are data in a predetermined format, and have a zero value. Converted into first data representing how many consecutive image data to be displayed or second data representing the value itself of image data representing a non-zero value and encoded The predetermined format is a first format applied when converting image data consisting of the difference between the conversion coefficients into the first data, and an image consisting of the difference between the conversion coefficients. A second format applied when converting data to the second data, each of the first format and the second format being a short format suitable for expressing a short bit length and a long Long format suitable for expressing bit lengthThe short format and the long format included in the first format include a code portion indicating 0 run length, and the short format and the long format included in the second format include a code portion representing the data itself. It is characterized in that the combination of the number of bits of the part is changed and the compression process is repeatedly performed, and the combination that maximizes the compression rate is selected.
[0010]
  According to the present inventionImage data decompression method isthe aboveAn image data decompression method for decompressing image data compressed using an image data compression method, comprising: decoding the compressed image data of each block according to the predetermined format; and And inversely wavelet transforming the block image data into one image data.
  According to this configuration, the data representing the 0 run length or the value of the image data itself may be the target of the decoding process. Therefore, decoding can be performed without performing complicated operations, and high-speed decoding is possible.
[0011]
  the aboveIn the image data decompression method, before the wavelet inverse transform, the difference between the image data is added, and then the quantized image data is inversely quantized.The
  the aboveIn the image data decompression method, the wavelet inversely transformed image data is converted from a YUV system to an RGB system.The
That is, an image data decompression method according to the present invention is an image data decompression method for decompressing image data compressed using the above image data compression method, and determines the format of the image data of each encoded block. And decoding the image data into image data comprising the difference between the transform coefficients, adding the difference between the transform coefficients of the decoded image data, and adding the difference between the transform coefficients. A step of inversely quantizing the obtained image data of each block, a step of inversely transforming the image data of each of the inversely quantized blocks, and image data of each of the blocks subjected to the inverse wavelet transform from the YUV system A step of converting to the RGB system, and a step of combining the image data of each block converted to the RGB system into one image data. Tsu including and up, the.
[0012]
  According to the present inventionThe image data decompression circuitthe aboveAn image data decompression circuit (e.g., an image data decompression circuit shown in FIG. 6 to be described later) that decompresses image data of each block compressed using an image data compression method. A decoder for decoding the image data of the block for each line in the horizontal direction or the vertical direction, and first and second registers for alternately storing the image data of each line decoded by the decoder A first wavelet inverse transform processing unit that alternately reads out image data of each line from the first and second registers and performs wavelet inverse transform processing; and the first wavelet inverse transform processing unit A buffer for storing image data of each line for one block, and the buffer in the other direction of the horizontal direction or the vertical direction. A second wavelet inverse transform processing unit for performing wavelet inverse transform processing on the image data stored in the memory, and a memory for storing image data of all blocks obtained by the conversion by the second wavelet inverse transform processing unit With.
[0013]
  the aboveIn an image data decompression circuit, an adder for adding the image data decoded by the decoder; and an addition result of the adder is dequantized, and the image data obtained by the inverse quantization is converted to the decoder An inverse quantizer for supplying the first and second registers as image data decoded byThe
  the aboveThe image data expansion circuit further includes a color display system conversion unit that converts the image data converted by the second wavelet conversion processing unit from YUV to RGB image data and supplies the image data to the memory.The
That is, an image data decompression circuit according to the present invention is an image data decompression circuit that decompresses image data of each block compressed using the above-described image data compression method. Decoding the image data for each line in the horizontal direction or the vertical direction by discriminating the format, and obtaining the image data consisting of the difference between the transform coefficients, and the image data obtained by the decoder An adder for adding a difference between transform coefficients formed, an inverse quantizer for inversely quantizing image data obtained as a result of the addition by the adder, and image data of each line obtained by the inverse quantizer. First and second registers that are alternately stored, and image data of each line are alternately read from the first and second registers, and wavelet inverse transform processing is performed. A wavelet inverse transform processing unit, a buffer for storing one line of image data of each line transformed by the first wavelet inverse transform processing unit, and the other direction of the horizontal direction or the vertical direction in the buffer A second wavelet inverse transformation processing unit for performing wavelet inverse transformation processing on the stored image data, and a color display system transformation for converting the image data converted by the second wavelet transformation processing unit from a YUV system to an RGB system And a memory for storing the image data obtained by the conversion by the color display system conversion unit.
[0014]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an image data compression method, decompression method, and image data decompression circuit according to embodiments of the present invention will be described in order with reference to the drawings.
A. Image data compression method
The image data compression method according to this embodiment will be described along the flow shown in FIG. 1 with reference to FIGS. In this image data compression method, the value of a component insensitive to human vision in the original image data to be compressed is replaced with “0”, and the appearance frequency of “0” is increased, thereby compressing the image data. Is.
[0015]
This will be specifically described.
First, as shown in FIG. 2, the original image data P to be compressed is divided into 12 blocks, block B0 to block B11 (step S11). In this embodiment, the original image data P is assumed to be RGB image data. Thereafter, as described below, a series of compression processing is performed for each block.
That is, the color display system of the image data of the first block B0 is converted from the RGB system to the YUV system (step S12). Here, the YUV color difference components (u, v) are components that are insensitive to human vision and have little effect on image quality even if they are thinned out. By thinning out, the processing speed is increased and the amount of data to be transmitted or stored is reduced.
[0016]
Subsequently, a wavelet transform process is performed in the horizontal direction of the block B0 on the image data of the block B0 whose color display system has been converted in the above-described step S12, and a horizontal direction component conversion coefficient is obtained (step S13). Thereafter, a wavelet transform process is performed in the vertical direction of the block B0 to obtain a transform coefficient of the vertical component (step S14). Here, the wavelet transform has a characteristic that it can extract image features. That is, according to the wavelet transform, the image data is separated into a low-frequency component sensitive to human vision and an insensitive high-frequency component. Among these, for the high frequency component, even if the conversion accuracy is lowered and data having a small absolute value is ignored, the influence on the image quality is small. In addition, since the value of the high frequency component subjected to the wavelet transform tends to be small, the appearance frequency of “0” further increases in the compression process.
[0017]
Subsequently, the horizontal and vertical transform coefficients of the image data of the block B0 obtained in the above steps S13 and S14 are quantized (step S15). Specifically, the effective bit length is shortened by rounding each transform coefficient by quantizing with appropriate accuracy. Thereafter, the conversion coefficients are arranged in descending order, and a conversion coefficient equal to or less than a predetermined threshold is set to “0” (threshold control). The above-described quantization accuracy and the predetermined threshold are experimentally determined in advance from the relationship between the required image quality and the compression rate. This threshold value is set as a changeable parameter, and can be arbitrarily changed.
[0018]
Subsequently, using the horizontal line as a unit, the conversion coefficient of the image data quantized in step S15 described above is scanned, and the difference in the conversion coefficient of the image data between adjacent pixels is calculated (step S16). That is, as the image data of each pixel, a difference with respect to the data of the adjacent pixel that was previously scanned is used. However, as the value of the conversion coefficient of the first pixel, the conversion coefficient obtained for that pixel is used. Here, in general, in the case of an image, since the frequency with which the spatial frequency is continuous is high, the values of adjacent pixels take values similar to each other, and the difference tends to decrease. For this reason, it is possible to suppress the number of data bits and reduce the amount of data by using the difference rather than using the absolute value of the conversion coefficient of each pixel.
[0019]
Subsequently, according to a predetermined format, the difference between the image data of the block B0 obtained in step S16 is encoded (step S17). In this format, the wavelet-transformed image data of each block includes 0 run length (first data) indicating how many consecutive image data indicating 0 value and image data value indicating non-zero value itself (first data). 2) is defined. FIG. 3 shows an example of a format used for encoding in step S17. In this embodiment, the format of the first data representing the 0 run length of the image data (hereinafter referred to as the first format) and the format of the second data representing the value of the image data itself (hereinafter referred to as the second format). Each format is further classified into a short format suitable for expressing a short bit length and a long format suitable for expressing a long bit length.
[0020]
FIG. 3A shows a short format of the first format, which is composed of a 2-bit header part and a code part of n (n is a natural number) bits. In this example, “00” is set in the header portion, and data encoded using the short format of the first format can be identified from the header portion. FIG. 3B shows a long format of the first format, which is composed of a 2-bit header part and an m-bit (m> n) code part. In this example, “01” is set in the header portion, and data encoded using the long format of the first format can be identified from the header portion.
[0021]
FIG. 3C shows a short format of the second format, which is composed of a 2-bit header portion and a k-bit (k is a natural number) bit code portion. In this example, “10” is set in the header portion, and from this header portion, data encoded using the short format of the second format can be identified. FIG. 3 (d) shows a long format of the second format, which is composed of a 2-bit header portion and an h-bit (h> k) code portion. In this example, “11” is set in the header portion, and from this header portion, data encoded using the long format of the second format can be identified.
[0022]
Furthermore, in this embodiment, as a modification of the first format described above, a third format indicating that all the remaining bits (all bits from the current focused bit to the last bit) are “0” is set. FIG. 3E shows an example of the third format. This third format has the same header portion as the short format of the first format described above, but has a code (all 0) consisting of n-bit “0” as the code portion. When data in the third format is detected during the decompression process, all subsequent bits are decoded as “0”.
[0023]
Here, the encoding using the format shown in FIG. 3 will be described with reference to FIG. However, the number of bits n, m, k, h in the format shown in FIG. 3 is n = 3, m = 5, k = 4, and h = 8. As the values of n, m, k, and h, the compression processing is actually performed according to various combinations, and the values having the highest compression rate may be used.
FIG. 4A shows an example of a data string to be encoded during the compression process. This data string is an array of the differences (hexadecimal display) of the conversion coefficients of each pixel in the block B0 obtained in step S16 described above. FIG. 4B shows an example of data obtained by encoding the data sequence shown in FIG. In this example, the long format of the second format is applied to the data portion “12” indicated by (1), and a 2-bit header part “11” and an 8-bit data string “12” itself are represented. It is encoded into data composed of a code part “00001100”.
[0024]
For the data portion “0” indicated by {circle around (2)} below, the short format of the first format is applied, indicating that the 2-bit header portion “00” and the 0 run length is “1”. It is encoded into data composed of a 3-bit code part “001”. The next short format of the second format is applied to the data portion “fe” (= −2) indicated by (3), and represents the 2-bit header portion “10” and the data “fe” itself. It is encoded into data composed of a 4-bit code part “1110”. For the data portion “ff” (= −1) shown in the next item (4), the short format of the second format is applied to represent the 2-bit header portion “10” and the data “ff” itself. It is encoded into data composed of a 4-bit code part “1111”. For the data portion “0 0 0 0 0 0 0 0” shown in the next item (5), the short format of the first format is applied, and the 2-bit header part “00” and the 0 run length are “7”. "Is encoded into data composed of a 3-bit code part" 111 "representing" ".
[0025]
The next short format of the second format is applied to the data portion “1” indicated by {circle around (6)}, a 2-bit header portion “10”, and a 4-bit code portion representing the data “1” itself. It is encoded into data composed of “0001”. The third data format is applied to the remaining data portion “0...” Shown in the next item (7), and is composed of a 2-bit header portion “00” and a 3-bit code portion “000”. Is encoded into the data to be processed. In this example, the data string shown in FIG. 4A is replaced with a bit string of 43 bits in total shown in FIG. 4B, and the image data is encoded.
As described above, the image data of the block B0 is compressed, and the image data is also sequentially compressed for the other blocks B1 to B11.
[0026]
B. Image data decompression method
In the following, a method (image data decompression method) for decompressing image data compressed using the above-described compression method will be described along the flow shown in FIG.
The image data of each block compressed using the compression method described above is decoded in accordance with the format shown in FIG. 3 used in the compression process (step S21).
[0027]
Here, decoding will be specifically described by taking image data including the bit string shown in FIG. 4B as an example. When decoding the bit string shown in the figure, the type of format is sequentially discriminated from the header part included in the bit string and adaptively decoded. In this example, the first two bits “11” of the bit string “1100001100” corresponding to the uppermost data “12” in FIG. 4B received first are extracted as a header part. Here, as shown in FIG. 3D, since the header part “11” represents the long format of the second format, the code part composed of the 8-bit bit string “00001100” following the header part “11” is displayed. Extracted as a bit string of the image data itself, a data portion “12” indicated by (1) in FIG.
[0028]
Subsequently, in the bit string “00001” corresponding to the data “0” shown in the second row from the top in FIG. 4B, the first 2 bits “00” are extracted as the header part. As shown in FIG. 3A, since the header part “00” represents the short format of the first format, the code part composed of the 3-bit bit string “001” following the header part “00” is used as the image data. Extracted as data representing 0 run length, data “0” indicated by (2) in FIG. 4A is obtained.
[0029]
Thereafter, similarly, the header part is sequentially extracted, and the code part is specified and decoded by referring to the format specified by the header part. The data decoded in this way corresponds to the difference between the transform coefficients of the image data obtained by performing the difference calculation in step S16 after being quantized in step S15 shown in FIG. However, the data of the first pixel corresponds to the conversion coefficient itself, not the difference. Note that the decoding process described above is performed in chronological order from the first compressed (encoded) data, and the header part and the code part are grasped for the decoded data. Therefore, even if the bit string forming the code portion matches the header portion, the position of the header portion of the data to be decoded next can be specified from the history of decoding of the data so far, and the decoding is performed. Will not interfere with the processing.
[0030]
Next, the difference between the conversion coefficients of the image data obtained by decoding is added to generate image data corresponding to each pixel (step S22). That is, the difference between the decoded transform coefficients is cumulatively added to generate a transform coefficient in units of horizontal lines. At this time, the decoded conversion coefficient is used as it is for the conversion coefficient of the first pixel in the horizontal direction, but the decoded value (difference) for each pixel is used as the conversion coefficient of each pixel thereafter. A value obtained by adding to the value of the previous pixel is used. Subsequently, the image data obtained by the addition in step S22 is inversely quantized (step S23). That is, with the horizontal line as a unit, the quantized value is multiplied by the transform coefficient obtained by addition in step S22 to return to the original bit length before being quantized in the compression process.
[0031]
Subsequently, the image data of each block decoded and inversely quantized as described above is subjected to wavelet inverse transform in the vertical direction (step S24), and then wavelet inverse transform is performed in the horizontal direction (step S25). Then, the color display system of the image data subjected to inverse wavelet conversion is converted from the YUV system to the RGB system (step S26), and the image data of each block converted to the RGB system is synthesized as one image (step S27). . Thus, the image data is expanded and one image can be reproduced.
[0032]
C. Image data decompression circuit
Next, an image data decompression circuit that decompresses image data according to the decompression method described above will be described. FIG. 6 shows the configuration of this image data decompression circuit. In the figure, the decoder 10 executes step S21 in the above-described decompression process, and inputs the image data (image data before decompression) compressed through the above-described steps S11 to S17. . The addition computing unit 20 executes step S22 in the above-described decompression process. The inverse quantizer 30 executes step S23 in the expansion process described above.
[0033]
The registers 41, 42, 43A and 43B store the inversely quantized transform coefficients. Among these, the register 41 stores the conversion coefficient of the color difference component u for one line in the vertical direction in the block. The register 42 stores the conversion coefficient of the color difference component v for one line in the vertical direction in the block. Further, each of the registers 43A and 43B stores a conversion coefficient for one line. That is, these registers can store the conversion coefficients of the color difference components u and v for one line and the conversion coefficient of the luminance component Y for two lines.
The color difference components u and v are thinned out in half in the horizontal and vertical directions, so the data amount of the color difference components u and v for one line is the luminance component for one line. It is half the data amount of Y.
[0034]
The wavelet inverse transform processing units 51, 52, and 53 execute step S24 in the above-described decompression process, and the conversion coefficients for one line in the vertical direction stored in the registers 41, 42, 43A, and 43B, respectively. Apply wavelet inverse transformation. Each of the buffers 61A, 61B, 62A, 62B, 63A, and 63B stores conversion coefficients for one block subjected to inverse wavelet transform. The wavelet inverse transform processing units 71, 72, 73 execute step S25 in the above-described decompression process, and transform one block stored in the buffers 61A, 61B, 62A, 62B, 63A, 63B, respectively. Wavelet inverse transform processing is applied to the coefficients in the horizontal direction.
[0035]
The color display system conversion unit 80 executes step S26 in the above-described decompression process, and converts the color display system of the image data obtained by the wavelet inverse transform from the YUV system to the RGB system. The memory 90 is used to execute step S27 in the above-described decompression process, and stores the decompressed image data of each block in association with the position of each block before compression. The memory 90 outputs image data of one finally expanded image.
[0036]
The operation (decompression method) of the image data decompression circuit will be described below with reference to FIGS. Note that the operations of the decoder 10, the adder 20, the inverse quantizer 30, the color display system converter 80, and the memory 90 shown in FIG. 6 are performed in the above-described steps S 21, S 22, S 23, S 26, S 27. Therefore, the description of these operations will be omitted. Here, the registers 41, 42, 43A, 43B, the wavelet inverse transform processing units 51, 52, 53, and the buffers 61A, 61B are omitted. , 62A, 62B, 63A, 63B, and each operation of the wavelet inverse transform processing units 71, 72, 73 will be described.
[0037]
7 shows the timing until the image data P0 of the block B0 before decompression in FIG. 6 is stored in the registers 41, 42, 43A, 43B, and the data stored in these registers is the buffers 61A, 61B, 62A, The timing until the data is stored in 62B, 63A, 63B is shown.
Of the conversion coefficients for one line of the image data Pn obtained by the decoder 10, the adder 20 and the inverse quantizer 30, the color difference component u is stored in the register 41, and the color difference component v is stored in the register. 42 is stored. The luminance component Y of the same line as these color difference components is stored in the register 43A or the register 43B. In this embodiment, since the color difference components u and v are thinned out for each line in the horizontal and vertical directions, two lines of the luminance component Y correspond to one line of the color difference components u and v. .
[0038]
That is, the image data P0 shown in FIG. 7 has the color difference components u0, v0 and the luminance component Y0 as the first line data, and only the luminance component Y1 as the second line data. Hereinafter, even-numbered lines have no color difference component. That is, even-numbered color difference components are thinned out. This is because attention is paid to the fact that even if a color difference component insensitive to human vision is thinned out, the influence on the image quality is small. That is, the amount of data to be transmitted or stored is reduced by thinning out color difference components that are insensitive to human vision.
[0039]
Next, the transform coefficients are read from the registers 41, 42, 43A, and 43B, respectively, and the vertical wavelet inverse transform of the first line is performed. More specifically, when the color difference components u0 and v0 of the first line are written in the registers 41 and 42 and the luminance component Y0 of the same line is written in the register 43A, the image data from these registers is subjected to wavelet inverse transform processing. The data are sequentially read by the units 51, 52, and 53, and the wavelet inverse transform process in the vertical direction of the first line is performed on each component. At this time, the wavelet inverse transformation process for the luminance component Y0 of the first line is performed in parallel with the process of writing the luminance component Y1 of the second line into the register 43B. Since the first line of data is read and processed, these processes are performed in parallel without any inconvenience. The color difference components u0, v0 and the luminance component Y0 of the first line obtained by performing the wavelet inverse transformation process are stored in the buffers 61A, 62A, 63A as image data Yuv0.
[0040]
Subsequently, the luminance component Y1 of the second line is read from the register 43B and subjected to wavelet inverse transformation processing by the wavelet inverse transformation processing unit 53. In this case, the color difference components are stored in the registers 41 and 42. The color difference components u0 and v0 in the first line are also used, and the wavelet inverse transform processing units 51 and 52 perform wavelet inverse transform processing on the color difference components u0 and v0. Each data obtained by performing wavelet inverse transform processing on the color difference components u0, v0 and the luminance component Y1 is stored in the buffers 61A, 62A, 63A as the image data Yuv1 of the second line. Similarly, the odd-numbered and even-numbered image data on and after the third line are inversely wavelet transformed in the horizontal direction, and the image data of the block B0 is stored in the buffers 61A, 62A, and 63A.
[0041]
Next, with reference to FIG. 8, the operation from reading out image data for one block stored in the buffers 61A, 62A, and 63A to performing horizontal wavelet inverse transformation processing and combining them into one image will be described. To do. The image data of the block B0 stored in the buffers 61A to 63A is read by the wavelet inverse transform processing units 71 to 73 and subjected to horizontal wavelet inverse transform processing, and then the RGB display is performed by the color display system transforming unit 80. And stored in the memory 90. Further, while the wavelet inverse transform processing units 71 to 73 are performing the wavelet inverse transform processing on the image data of the block B0, the above-described decoder 10, addition calculator 20, inverse quantizer 30, register 41 , 42, 43A, 43B, and the wavelet inverse transform processing units 51, 52, 53 execute the above-described steps S21 to S24, and the image data of the block B1 is written into the buffers 61B to 63B.
[0042]
The image data written in the buffers 61B to 63B is read into the wavelet inverse transform processing units 71 to 73 after the horizontal wavelet inverse transform processing for the image data stored in the buffers 61A to 63A is completed. Wavelet inverse transform processing is performed. Similarly, the color display system is converted and stored in the memory 90. Similarly, the image data of the odd-numbered and even-numbered blocks are alternately read out from the buffers 61A to 63A and the buffers 61B to 63B, and the wavelet inverse transformation in the vertical direction is performed, so that the color display system is It is converted and stored in the memory 90. Thus, the image data of all blocks has been expanded.
[0043]
According to this image data decompression circuit, two registers 43A and 43B for storing a total of two lines of luminance components Y are provided in the preceding stage of the wavelet inverse transform processing unit 53, so that the luminance components of each line within the same block. The wavelet inverse transform process in the vertical direction with respect to Y can be pipelined. Further, since two systems of buffers 61A, 62A, 63A and buffers 61B, 62B, 63B are provided in the previous stage of the wavelet inverse transform processing units 71, 72, 73, the horizontal wavelet inverse transform processing for each block is pipelined. Can be Therefore, it is possible to speed up the decoding process. The image data decompression circuit shown in FIG. 6 has been described above.
[0044]
The structural features of this embodiment will be described below.
(1) When image data is converted using wavelet transform, the characteristics of the image can be extracted.
(2) Even if the accuracy of wavelet-transformed data is reduced or a small absolute value is ignored, the influence on the image quality is small, so the number of bits of image data can be reduced or 0 can be replaced with continuous data. . Thereby, 0 run length is improved and a compression rate improves. In addition, if the color display system of the image data is a YUV system, the compression rate is further improved.
(3) If it is desired to prioritize the compression rate even at the expense of image quality, significant compression is possible by adjusting the accuracy of the wavelet transformed data.
[0045]
(4) In encoding, a method is used in which the difference between data of adjacent pixels is calculated to further increase the appearance frequency of 0 and improve the compression rate. Further, in decoding, the difference only needs to be added, and no complicated calculation is required, so that the decoding process can be speeded up.
(5) The image data decompression circuit performs pipeline processing. For this purpose, a register that holds image data in units of one line in the inverse wavelet transform in the vertical direction (horizontal direction) and a buffer that holds data for one block in the wavelet inverse transform in the horizontal direction (vertical direction) Each has a double. Further, the number of pipeline stages may be increased by multiplexing.
[0046]
The effects of this embodiment will be described below.
(1) Since encoding is performed using two types of effective bit length formats and two types of 0-run length formats, a compression / image data expansion circuit can be provided with a simple circuit configuration without requiring a table or complicated calculation. Can be realized.
(2) When wavelet transform is performed for all pixels of one image, a large-capacity buffer is required. According to this embodiment, an image is divided into a plurality of blocks of an appropriate size. Since processing is performed, the capacity of the buffer for storing image data for one block can be reduced. Moreover, by dividing, pipeline processing becomes possible, and the decoding process can be speeded up.
[0047]
Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and design changes and the like within a scope not departing from the gist of the present invention are included in the present invention. . For example, in the above-described embodiment, the wavelet transform in the horizontal direction is performed after the wavelet transform in the horizontal direction in the compression process. Conversely, the process in the horizontal direction is performed after performing the conversion process in the vertical direction. In the decoding process, the order of the conversion process in the vertical direction and the conversion process in the horizontal direction may be interchanged. Further, although the color display system of the original image data P is an RGB system, it may be a YUV system. In this case, color display conversion processing is not required in the compression processing and decompression processing. Furthermore, as long as the required compression rate can be obtained, the quantization / inverse quantization, threshold control, and difference / addition operations may be omitted.
[0048]
【The invention's effect】
As described above, according to the present invention, in the compression process, the image data to be compressed is divided into a plurality of blocks, subjected to wavelet transform, and the image data of each block subjected to the wavelet transform is encoded according to a predetermined format. In the decompression process, the compressed image data of each block is decoded according to the predetermined format and inverse wavelet transform is performed, so that the image data is decompressed at high speed while suppressing the influence on the image quality. Can be realized, and hardware implementation becomes easy.
[Brief description of the drawings]
FIG. 1 is a flowchart showing a flow of image data compression processing according to an embodiment of the present invention.
FIG. 2 is a diagram showing an example of block division of image data according to the embodiment of the present invention.
FIG. 3 is a diagram for explaining a data format used in compression and decompression processing according to the embodiment of the present invention;
FIG. 4 is a diagram showing an example of encoding according to the embodiment of the present invention.
FIG. 5 is a flowchart showing a flow of image data decompression processing according to the embodiment of the present invention.
FIG. 6 is a block diagram showing a configuration of an image data expansion circuit according to the embodiment of the present invention.
FIG. 7 is a diagram for explaining the operation of the image data decompression circuit according to the embodiment of the present invention (the image data generation process of each block).
FIG. 8 is a diagram for explaining the operation of the image data expansion circuit according to the embodiment of the present invention (the process of synthesizing the image data of each block).
[Explanation of symbols]
S11 to S17; Step (compression processing), S21 to S27; Step (decompression processing), 10; Decoder, 20; Addition calculator, 30; Inverse quantizer, 41, 42, 43A, 43B; Register, 51 , 52, 53; Wavelet inverse transformation processing unit (vertical direction), 61A, 61B, 62A, 62B, 63A, 63B; Buffer, 71, 72, 73; Wavelet inverse transformation processing unit (horizontal direction), 80; Color display system Conversion unit, 90; memory.

Claims (6)

A first step of dividing image data to be compressed into a plurality of blocks;
A second step of converting the divided image data from an RGB system to a YUV system;
A third step of wavelet transforming the image data converted into the YUV system for each block, and thinning out the color difference component at that time
A fourth step of quantizing the transform coefficient of the image data obtained by the wavelet transform;
A fifth step of calculating a difference between the quantized transform coefficients and transform coefficients between adjacent pixels;
The image data composed of the difference between the conversion coefficients is data of a predetermined format, and represents the first data indicating how many consecutive image data indicating 0 value or the value of image data indicating non-zero value itself. A sixth step of converting and encoding the second data,
The predetermined format is:
The first format applied when converting the image data consisting of the difference between the conversion coefficients into the first data, and applied when converting the image data consisting of the difference between the conversion coefficients into the second data. Second format,
Said first respective format and the second format, and short format suitable for representing a short bit length, viewed it contains a long format suitable for representing bit length, included in the first format The short format and the long format include a code part indicating 0 run length, the short format and the long format included in the second format include a code part representing the data itself, and the combination of the number of bits of each code part is changed. An image data compression method comprising: repeatedly performing compression processing and selecting a combination that maximizes the compression rate .   2. The image data compression method according to claim 1, wherein, in the third step, color difference components are thinned out during the wavelet transform.   3. The image data compression method according to claim 1, wherein in the fourth step, a conversion coefficient equal to or less than a predetermined threshold value is set to 0 value. 4.   4. Each of the short format and the long format includes a 2-bit header portion for identifying the format of the data encoded in the sixth step. The image data compression method described in the item. An image data decompression method for decompressing image data compressed using the image data compression method according to claim 1,
Determining the format of the encoded image data of each block and decoding the image data into image data comprising the difference of the transform coefficients;
Adding a difference of transform coefficients of the decoded image data;
Dequantizing the image data of each block obtained by adding the difference of the transform coefficients;
Wavelet inverse transform of the inverse quantized image data of each block;
Converting the wavelet inversely transformed image data of each block from YUV system to RGB system;
Combining the image data of each block converted into the RGB system into one image data;
A method for decompressing image data. An image data decompression circuit that decompresses image data of each block compressed using the image data compression method according to claim 1,
A decoder that determines the format of the image data of each of the encoded blocks, decodes the image data for each line in the horizontal direction or the vertical direction, and obtains image data including a difference between the transform coefficients;
An adder for adding a difference of transform coefficients forming the image data obtained by the decoder;
An inverse quantizer that inversely quantizes image data obtained as a result of the addition by the adder;
First and second registers for alternately storing image data of each line obtained by the inverse quantizer;
A first wavelet inverse transform processing unit that alternately reads out image data of each line from the first and second registers and performs wavelet inverse transform processing;
A buffer for storing the image data of each line converted by the first wavelet inverse transform processing unit for one block;
A second wavelet inverse transform processing unit for performing wavelet inverse transform processing on the image data stored in the buffer with respect to the other of the horizontal direction and the vertical direction;
A color display system conversion unit for converting the image data converted by the second wavelet conversion processing unit from a YUV system to an RGB system;
A memory for storing image data obtained by conversion by the color display system conversion unit;
An image data decompression circuit.

Images