JP2010219798A - Image encoding/decoding device, image encoding/decoding method, program, recording medium, and image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an encoding/decoding device that suppresses deterioration in image quality and executes high-speed processing by effectively utilizing idle bits for flags in order to accurately execute addition of r by accurately determining a bit plane to be a truncation point. <P>SOLUTION: The encoding/decoding device performs encoding/decoding by processing image data while using the JPEG 2000. The device includes a fixed-point arithmetic means for executing fixed-point arithmetic processing for subjecting the image data to arithmetic processing, a 9×7 filter means used for arithmetic processing to execute Wavelet transform of the image data, a scalar quantization means for executing scalar quantization processing to quantize the image data, an entropy encoding means for executing entropy encoding processing to encode/decode the image data, and a data buffering means for storing image data subjected to arithmetic processing. The data buffering means uses the fixed number of bits for storage of the data subjected to the arithmetic processing while utilizing idle bits of a data buffer for decoding processing. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an image encoding / decoding device, an image encoding / decoding method, a program, a recording medium, and an image processing device.

  As a next-generation high-efficiency encoding system for image data, ISO (International Organization for Standardization) and ITU-T (International Telecommunication Union Telecommunication Standardization Sector) have developed the JPEG2000 (Joint Photographic Experts Group 2000) system. The JPEG2000 system has superior functions compared to the current mainstream JPEG (Joint Photographic Experts Group) system, and adopts DWT (Discrete Wavelet Transform) as an orthogonal transform, and bit for entropy coding. It is characterized in that a method called EBCOT (Embedded Block Coding with Optimized Truncation) for performing plane coding is adopted.

  In JPEG 2000 Part I (ISO / IEC 15444-1), if the value of each quantized wavelet coefficient is not “0” during the decoding process, an arbitrary r (0 ≦ r <1) is set at the lowest order. It is prescribed to add. This is to mitigate the effect of rounding to reduce the absolute value of the wavelet coefficient in the quantization during the encoding procedure.

  For example, FIG. 2 shows each wavelet coefficient of “16, 64, 88, 32, 0, 0, 0, 112” converted into bit planes, where “the number of bits is 9” and “the number of 0 bit planes is 2. ”And“ the number of processing passes is 3 ”. These three paths are three paths called “Significant Propagation Pass”, “Magnitude Refinement Pass”, and “Cleanup Pass”, and the boundary of each path is a truncation point. A bit plane divided by three paths is called a sub-bit plane. The values of all bit data below the truncation point are “0”, and the processing is substantially the same as that of truncation. JPEG2000 intends to perform compression encoding as efficiently as possible by such truncation processing.

  Conventionally, decoding processing is performed by adding r to the lowest order after inverse quantization.

  For example, in Patent Document 1, the value of the parameter r corresponding to the attention coefficient is determined based on the subband to which the quantized attention coefficient belongs and the coefficient in the vicinity of the attention coefficient, and the determined parameter is quantized. It is disclosed that it is added to each normalized wavelet coefficient. In this method, a truncation point is determined for each bit plane and an addition process is performed. However, in the case of the truncation point for each pass unit, there is a problem that the bit plane of the truncation point cannot be accurately determined and the image quality is somewhat deteriorated.

  On the other hand, conventionally, a processing method for speeding up has been proposed. For example, as described in Japanese Patent Application Laid-Open No. H10-228561, there is a technique that speeds up by reducing path processing.

  Further, as another method, there is a method of adding r in units of sub bit planes. Specifically, decoding is performed on a sub bit plane to be decoded, and r is added. When the decoding target sub-bit plane moves to the next time, the decoding process is performed by subtracting and decoding r added in the above, and adding r to the target sub-bit plane. I think that. This requires the addition and subtraction processing of r for all the sub-bit planes, resulting in a problem of a reduction in processing speed.

  Therefore, it is possible to accurately determine a bit plane serving as a truncation point, and to suppress deterioration in image quality in decoding processing by adding an arbitrary r to lower bit data than the truncation point using the bit plane. Also, by minimizing the arbitrary r addition processing, a high-speed processing method can be achieved.

  However, in the case of the above method, for example, when the truncation point is bit plane 0, the addition process of r is not performed. When the r value is ½, when the truncation point is bit plane 0, the addition processing is ½, and addition after the decimal point also suppresses image quality degradation. However, there is a problem that the addition processing is not performed because the coefficient buffer of the processing portion that performs the addition processing does not secure the decimal point.

  Therefore, in order to perform the addition processing after the decimal point, if an empty bit is created in the coefficient buffer and the addition processing after the decimal point is necessary, a flag is added to the empty bit, and the addition processing after the decimal point is performed in the post-processing. To do.

  Therefore, the present invention has been made in view of the above problems, and performs addition of r that accurately determines a bit plane that is a truncation point, and also performs addition accurately when r is a decimal point or less. Therefore, it is an object of the present invention to suppress image quality degradation, perform high-speed processing, and effectively use a memory by effectively using an empty bit for a flag.

  In order to solve the above-described problems, an encoding / decoding device according to the present invention processes image data by JPEG2000, which is an encoding method for encoding image data, and encodes / decodes the image data. Fixed decimal arithmetic means for performing fixed decimal arithmetic processing for arithmetic processing, 9 × 7 filter means used for arithmetic processing for performing wavelet transform of image data, and a scalar that is arithmetic processing for quantizing image data Scalar quantization means for performing quantization processing, entropy encoding means for performing entropy encoding processing that is processing for encoding / decoding image data, and data processed to perform image data processing Data buffer means for storing data, and the data buffer means stores a fixed number of bits for storing the processed data. Used, characterized by utilizing the empty bits of the data buffer in the decoding process.

  The encoding / decoding method according to the present invention is a method for processing image data by JPEG2000, which is an encoding method for encoding image data, and a fixed method for calculating image data. A fixed decimal arithmetic step for performing decimal arithmetic processing, a 9 × 7 filter step used for arithmetic processing for performing wavelet transform of image data, and a scalar for performing scalar quantization processing that is arithmetic processing for quantizing image data The entropy encoding process that performs the quantization process, the entropy encoding process that is an operation process for encoding / decoding the image data, and the storage of the data processed to perform the image data calculation process are fixed. And a data buffer process for using empty bits of the data buffer for decoding processing using the number of bits.

  The program according to the present invention is a program that processes and encodes and decodes image data according to JPEG2000, which is an encoding method for encoding image data, and performs fixed decimal arithmetic processing for calculating image data. Fixed decimal arithmetic processing to be performed, 9 × 7 filter processing used for arithmetic processing for performing wavelet transform of image data, and scalar quantization processing to perform scalar quantization processing that is arithmetic processing for quantizing image data; A fixed number of bits is used to store the entropy encoding process for performing the entropy encoding process, which is an arithmetic process for encoding / decoding the image data, and the data processed for the image data calculation process. , Causing the computer to execute data buffer processing for using empty bits of the data buffer for decoding processing It is characterized in.

  A computer-readable recording medium according to the present invention records a program for executing the encoding / decoding method described above.

  According to the present invention, it is possible to perform high-speed image encoding / decoding processing in which deterioration of image quality is suppressed.

It is a block diagram which shows the structure of the conventional image decoding apparatus using JPEG2000. It is a figure explaining post-quantization. It is a figure explaining the bit plane formed by slicing a code block for every bit. It is a figure explaining the scanning order at the time of decoding the coefficient on a bit plane. It is a figure explaining eight neighboring coefficients adjacent to the circumference | surroundings of the coefficient X of the decoding object on a bit plane. It is a figure explaining the classification table of the context in S pass. It is a figure explaining the vertical parameter | index and horizontal parameter | index table for determining the context of a polarity bit. It is a figure which shows the table which matched the combination of the horizontal parameter | index h and the vertical parameter | index v, and the context label CX with respect to five types of polarity bits. It is a figure which shows the classification table of the context in R path | pass. It is a figure which shows the classification table of the context in C path | pass. It is a figure which shows the flow about an entropy decoding means. It is a figure which shows the flow about an entropy decoding means. It is a figure which shows the state by which code data was divided | segmented into the sub bit plane. It is a figure which shows the state by which code data was divided | segmented into the sub bit plane. It is a figure which shows the state by which code data was divided | segmented into the sub bit plane. It is a figure which shows the state by which code data was divided | segmented into the sub bit plane. It is a figure which shows the flow about the flag insertion to an empty bit. It is a figure which shows the flow about the addition process of r value which uses the flag of an empty bit. 1 illustrates an example of a device configuration according to an embodiment of the present invention.

  First, before describing the details of the embodiment, a description will be given of empty bits. The 9 × 7 wavelet is a real number operation, and the wavelet coefficient is originally handled as a floating point, but here it is handled as an integer (fixed decimal) in order to speed up the calculation. The bit configuration of the 9 × 7 wavelet coefficient is an original image 8 bits input at the time of encoding, and the 9 × 7 wavelet coefficient is a real number. Therefore, it is necessary to store both the integer part and the decimal part in the coefficient buffer. The number of bits increased by the wavelet transform is LL1 bit, HL2 bit, LH2 bit, and HH3 bit. Therefore, the maximum number of bits after wavelet transformation is LL9 bits, HL10 bits, LH10 bits, and HH11 bits. Since the data buffer in the computer normally uses 8, 16, 32, and 64 bits, and the integer part needs 11 bits, 16 bits are used for the data buffer. The decimal part needs to be close to a floating point in order to improve the PSNR, and the remaining 5 bits are used to increase the number of digits of the decimal part as much as possible. Therefore, there is no free bit here.

  Thereafter, linear quantization and integerization are performed on the 16-bit coefficient. The linear quantization in JPEG2000 is actually normalization, and in the case of most codecs, there is a special situation that division is performed with the same number of quantization steps.

  Next, a method for calculating the number of quantization steps that maximizes the PSNR under a predetermined compression rate will be described. The number of quantization steps that maximizes the PSNR in the case of a monochrome image will be described. Derivation relating to the number of linear quantization steps in non-orthogonal transformation is described in Non-Patent Document 1 and Non-Patent Document 2.

  Non-Patent Documents 1 and 2 derive conditions for giving an optimum PSNR by Lagrange's undetermined multiplier method (a method of obtaining a desired optimal solution by matching a partial differential value with a value of an undetermined multiplier), All other literature published later follows this approach. In this method, the quantization error can be minimized by setting the distortion slope (Δ quantization error / Δ code amount) corresponding to the Lagrange's undetermined multiplier λ to the same value for each subband.

  Let Gsb be the energy gain (hereinafter referred to as subband gain) of inverse wavelet transform in a certain subband sb, and let Δsb be the number of quantization steps applied to that subband. After going through several approximations for quantization error, the distortion slope is finally given as a constant multiple of “Gsb · Δsb ^ 2”. If this value is the same in each subband, quantization is performed under a predetermined code amount. The error is minimized (Non-Patent Documents 1 to 4).

  Therefore, in order to maximize the PSNR, the number of quantization steps to be applied to the subband is given by Equation (1). Here, r is a constant determined by the desired code amount, and the representative value is 1. In this embodiment, it is assumed that the representative value 1 is adopted.

  Next, the subband gain of the subband gain calculation method will be described. In inverse transform of frequency transform that is not orthonormal transform, the “magnification when transform coefficient value is inversely transformed into signal value” differs for each frequency band, that is, subband, and the square of this magnification is called subband gain. The error Δe generated in the transform coefficient due to the quantization is multiplied by the square root of the subband gain by the inverse transform to the signal, and is given by Equation (2).

  Hereinafter, for the sake of simplicity, description will be made by taking a 5 × 3 wavelet as an example. When x is a coordinate value, the inverse wavelet transform of JPEG2000 applies an inverse low-pass filter to the array of interleaved coefficient C (x), centering on a coefficient where x is an even number (x = 2i) in the horizontal direction. Next, an inverse high-pass filter is applied around the coefficient where x is an odd number (x = 2i + 1). Here, the inverse low-pass filter and the inverse high-pass filter are expressed by the following equations (3) and (4) in order.

  Taking the floor functions of equations (5) and (6), the following approximate expression is obtained.

  Further, the following five equations are obtained from equations (5) and (6).

  Now, when a quantization error 1 occurs in the high-pass coefficient C (2i + 1) at odd positions, the above five expressions (7) to (11) have the error from P (2i-1) to P (2i + 3). ) To affect 5 pixels.

  Assuming that these five errors are independent, the square root of the sum of squares of errors occurring in the five pixels is expressed by the following equation (12).

  That is, the high-pass coefficient error 1 is converted into a pixel value error 0.85. Here, −1/8, −1/4, 3/4, −1/4, and −1/8 are constants of the low-pass filter at the time of forward conversion, and in general, the sum of these squares is equivalent to one inverse high-pass filter. A subband gain GH is given.

  Similarly, when the quantization error 1 occurs in the low-pass coefficient C (2i) at the even position, the above five equations (7) to (11) have the error from P (2i-1) to P ( 2i + 1) 3 pixels are affected. Therefore, the square root of the sum of squares of errors generated in the three pixels is expressed by Equation (13). In general, the square sum of the constants of the high-pass filter at the time of forward conversion gives the subband gain GL for one inverse low-pass filter.

  In addition, the subband gain when the inverse transformation A is applied in the horizontal direction and then the inverse transformation B is further applied in the horizontal direction is the product of the subband gains of the inverse transformations a and B. That is, the subband gain of the HL coefficient is GH · GL.

  Similarly, when performing inverse conversion from the decomposition level n to n-1, for example, the subband gain of the (n-1) HL coefficient is obtained by multiplying the subband gain of the nLL coefficient by the subband gain of the HL coefficient. Become.

  Thus, in general, the subband gain is given by:

The nLL coefficient is (GL ^ 2) ^ n,
The nHL coefficient and the nLH coefficient are (GL · GH) · (GL ^ (n-1)),
The nHH coefficient is (GH ^ 2) · (GL ^ (n-1)).

  Tables 1 and 2 show the square roots of the subband gains up to 5 × 3 and 9 × 7 decomposition levels 5 below. Table 1 represents the square root of the subband gain of the 5 × 3 inverse wavelet transform. Table 2 represents the square root of the subband gain of the 9 × 7 inverse wavelet transform.

  Further, the fact that the number of quantization steps for each subband is expressed by equation (14) means that the magnification of the error caused by truncation after linear quantization to the signal value is made equal for each subband. Therefore, the linear quantization in JPEG2000 is actually normalization.

  As described above, if the wavelet transform is determined, the subband gain is calculated as a unique constant, and the quantization step number is determined based on the constant. Considering the fact that the representative value of r is 1, it can be seen that the number of quantization steps in JPEG 2000 can be handled as a substantially constant.

  The number of quantization steps that maximizes the PSNR in the case of a color image will be described. In the case of color, it is necessary to add the gain Gct of inverse component conversion to the subband gain Gsb, and the number of quantization steps that maximizes the PSNR is given by Expression (15). Here, Gct is constant regardless of the subband, and s is a constant similar to (1.1). However, when the gain of luminance Y is GctY, it is normal to use Expression (16). Similarly, the gain of the color difference Cb is GctCb, and the gain of the color difference Cr is GctCr. That is, even in the case of a color image, the same quantization step number as that in the case of a monochrome image is used. For Cb and Cr, the ratio of Expression (18) to Expression (19) with respect to Expression (17) is used. It is converted.

  The gain of inverse conversion of irreversible color transform (irreversible component transform, ICT) will be described. First, inverse ICT is given by the following equation, and is used simultaneously with inverse transformation of an irreversible 9 × 7 wavelet. On the other hand, inverse RCT is used simultaneously with inverse transformation of a reversible 5 × 3 wavelet.

  Table 3 shows the definition of inverse ICT. From the above equation in Table 3, for example, when a unit quantization error occurs in the color difference Cb, the square root of the square sum of the quantization error generated in the RGB value by inverse ICT is given by the following equation 20.

  Thus, Table 4 below is obtained. Table 4 represents the square root of the inverse ICT gain.

  Based on the above, considering the bit depth of each subband, the number of bits increases due to linear quantization, but the coefficient is converted to an integer for bit plane conversion, so it is necessary to store the decimal part in the coefficient buffer. Disappear. The number of bits increased by quantization is LL5bit, HL4bit, LH4bit, HH3bit, and the maximum number of bits after quantization is LL14bit, HL14bit, LH14bit, HH14bit.

  That is, at the time of decoding, the maximum number of bits of the coefficient buffer input before decoding is LL14bit, HL14bit, LH14bit, HH14bit. As a result, when the original image is 8 bits for all subbands, the upper 2 bits are vacant in order to secure 16 bits for the coefficient buffer. This is a result of the special circumstances of fixed decimal arithmetic and quantization for normalization purposes.

  The reason why these high-order bits after quantization are usually left free is that when they are handled in other than the normal data type (8, 16, 32, 64 bits), every time one coefficient is accessed, the bit unit is used. This is because the operation is accompanied, and contrary to the speed increase, the buffer cannot be made small. Here, considering reconstruction, a flag is necessary, but it is uneconomical to hold a memory for the flag.

  Next, the present embodiment will be described in detail with reference to the drawings.

<Configuration of First Embodiment>
FIG. 1 is a block diagram showing a configuration of a conventional image decoding apparatus using JPEG2000. This apparatus includes code data input means 1, header analysis means 2, entropy decoding means 3, scalar inverse quantization means 4, inverse wavelet transform means 5, color conversion means 6, DC level shift means 7, and image data output means. 8 is provided. Of these, the scalar inverse quantization means 4, the color conversion means 6, and the DC level shift means 7 are provided as options in JPEG2000. There are other options in JPEG2000, but they are not directly related to the present embodiment, or even if they are present, the explanation is omitted.

  Next, the operation of FIG. 1 will be described. First, image data compression-encoded according to the JPEG2000 standard is input by the code data input means 1 using a file system or network means. This image data is provided with a main header and a tile header, and the header analysis means 2 reads and analyzes these headers. Here, the main header stores information about the size of the image, the size of the tile, the offset of the image position, the offset of the tile position, the number of sample bits per pixel, the number of components such as RGB, and the like. . The tile header stores a tile number, a tile size, and the like, and can specify an encoding style, a quantization style, and the like.

  The entropy decoding unit 3 performs entropy decoding of the compression-encoded image data using the analysis result of the header analysis unit 2, and creates a scalar quantized coefficient sequence from the output bit sequence.

  The scalar inverse quantization unit 4 converts the coefficient sequence created by the entropy decoding unit 3 into a predetermined number in the JPEG 2000 standard based on the quantization style specified by the main header and the tile header analyzed by the header analysis unit 2. Wavelet coefficients are generated by inverse quantization using a calculation formula. Note that the wavelet coefficients generated by the wavelet transform are not necessarily quantized at the time of encoding, and in this case, the inverse quantization by the scalar inverse quantization means 4 is not performed.

  The inverse wavelet transform unit 5 performs inverse wavelet transform on the wavelet coefficient obtained by the scalar inverse quantization unit 4 based on the number of decompositions of the coding style specified by the main header or tile header, and converts the image data in the YUV color space. create.

  The color conversion unit 6 performs color conversion of the image data obtained by the inverse wavelet conversion unit 5 from the YUV component to the RGB component according to the analysis result of the header analysis unit 2. However, the color conversion means 6 does not operate when code data of a monochrome image or color conversion is not performed at the time of encoding.

  The DC level shift means 7 shifts the DC level of each value of the RGB components obtained by the color conversion means 6. The DC level shift is a shift so that the dynamic range is centered on the value 0 at the time of encoding, and is returned to the original at the time of decoding. For example, assuming that certain input image data is represented by 8 bits, this data takes any value from 0 to 255. By shifting the DC level, the value of this data is set to 0. A value of -128 to 127 can be set as the center. As described above, the reason for performing the DC level shift is that the coefficient after frequency conversion is encoded by being divided into a positive and negative sign and an absolute value, and therefore the data arrangement is based on the zero value as a reference. This is because the efficiency is increased. The image data output means 8 is a means for outputting the image obtained by the DC level shift means 7.

  In JPEG2000, data truncation processing called “post-quantization” (sometimes referred to as “truncation”) is performed on image data after wavelet transform and coefficient bit modeling. But not always.

  FIG. 2 is an explanatory diagram of this post-quantization. Shown in this figure are some bit-plane wavelet coefficients included in one code block. The header information called packet header information based on the JPEG2000 standard stores data such as the number of bits of the wavelet coefficient, the number of zero bit planes, the number of processing paths, and the like.

  In JPEG2000, the amount of codes is reduced by post-quantization, and efficient image compression is performed. However, when post-quantization is performed, all the lower-order bit data from the truncation point is uniformly zero, so that when the data is decoded and played back as it is, the reproduced image includes a part that is uncomfortable or unclear. The image quality may be degraded. Therefore, by adding an arbitrary r to the lower-order bit data below the truncation point, image quality degradation of the decoding process is suppressed. Conventionally, the point at which arbitrary r is added is a truncation point in bit plane units. However, when the truncation point is in units of each path, the bit plane that becomes the truncation point cannot be determined.

  Therefore, the present invention accurately determines a bit plane serving as a truncation point, and suppresses deterioration of image quality in decoding processing by adding an arbitrary value r to lower bit data than the truncation point using the bit plane. By minimizing the arbitrary r addition process, a high-speed processing method is obtained.

  A detailed flow of the operation of FIG. 1 will be described. The header analysis means 2 identifies the head address of the packet data by decoding the COD marker. Based on the head address of the identified packet data, the encoded data CD of the first encoding pass is taken out, and the S pass, R pass, and C pass encoded streams are sequentially output to the entropy decoding means 3. . The entropy decoding unit 3 receives the encoded data CD and the context label CX as inputs, and sequentially decodes each path in the order of the input paths. The decision D output from the entropy decoding means 3 is output to the scalar inverse quantization means 4.

  The context label CX is determined by performing a coefficient bit modeling process at the time of decoding in JPEG2000. The coefficient bit modeling process at the time of decoding is a process of determining the context required for the arithmetic decoding while updating the peripheral information using the decoding result by the entropy decoding means 3. Hereinafter, the coefficient bit modeling process at the time of decoding will be described with reference to FIGS.

  The subband image extracted from the encoded stream by the header analysis unit 2 is divided into small areas called code blocks, and the code block is a decoding processing unit. The subband image has an LL subband having a low frequency component in both the x and y directions of the original image, a low frequency component in one of the x and y directions, and a high frequency component in the other direction. The HL and LH subbands are either HH subbands having high frequency components in both the x and y directions.

  The size of the code block is an integer value of a power of 2 in which the height and width are in the range of 22 to 210, and the standard defines that the sum of the height and width indices is 12 or less.

  FIG. 3 is a diagram for explaining a bit plane formed by slicing a code block for each bit. The bit plane of the code block is composed of a Sign plane that stores the sign of the quantized value of the wavelet transform coefficient, and a plurality of planes from MSB to LSB that give the absolute value of the quantized value of the wavelet transform coefficient. . The entropy decoding means 3 selects bit planes in order from the upper bits, and performs decoding processing of each coefficient on the bit plane in units of bit planes.

  FIG. 4 is a diagram for explaining the scanning order when the coefficients on the bit plane are decoded. The bit plane is divided into stripes of 4 pixels in length, and is scanned and decoded in units of stripes. The width of each stripe is the width of the code block, which is 32 pixels in this example. The total number of stripes is 8.

  In each stripe, as indicated by the arrows, the vertical four pixels in the first column are scanned in order from the top, and the decoding process proceeds one pixel at a time, and then the vertical four pixels in the second column are scanned in the same direction. Repeat until the last column. When the decoding process is completed up to the end of the stripe, scanning is performed in the same manner from the leading coefficient of the next stripe, the decoding process is performed, and the process is repeated until the last line of the code block. Note that the decoding of the Sign plane storing the positive and negative signs of the coefficients is appropriately performed when decoding the plane corresponding to the absolute value portion of the coefficients.

  Each coefficient in the code block is identified in two states: significant or insignificant. 1 is assigned if significant, 0 is assigned if not significant, and the significance state of each coefficient is identified by a binary value. At the start of decoding, all coefficients in the code block are not significant.

  Here, the coefficient is significant means that the quantized value of the focused wavelet transform coefficient is known to be not “0” from the results of the decoding processing so far. In other words, it means the state of the wavelet transform coefficient in which a bit of “1” is found while decoding while scanning in units of stripes sequentially from the upper bit plane. Further, “the coefficient is not significant” means a state where the quantized value of the focused wavelet transform coefficient is “0” or remains “0”. In other words, it means the state of the wavelet transform coefficient in which the bit of “1” has not yet been found during the sequential decoding from the upper bit plane.

  In the coefficient bit modeling process, the context is determined for each coefficient bit based on the significance state of the coefficients around the coefficients on the bit plane.

  FIG. 5 is a diagram illustrating eight neighboring coefficients adjacent to the periphery of the coefficient X to be decoded on the bit plane (hereinafter referred to as the current coefficient X). The eight neighboring coefficients are two coefficients h0 and h1 adjacent to the current coefficient X in the horizontal direction, two coefficients v0 and v1 adjacent in the vertical direction, and four coefficients d0 to d3 adjacent in the oblique direction. It consists of.

  Since each of the eight neighboring coefficients in FIG. 5 can take two states, which are significant or not significant, there can be 256 contexts as a combination with respect to the current coefficient X. It is collected into 19 kinds of contexts using the sex etc., and these contexts are identified with labels of 0-18. If any neighborhood coefficient is located outside the code block being processed, the neighborhood coefficient is regarded as insignificant and processed.

  Each coefficient bit in the bit plane is decoded based on one of three types of processing paths, namely, an S path (significance propagation pass), an R path (magnitude refinement pass), and a C path (cleanup pass). One bit plane is divided into three sub bit planes as three types of processing paths. In the S pass, a non-significant coefficient around which a significant coefficient exists is decoded, in the R pass, a significant coefficient is decoded, and in the C pass, the remaining coefficients are decoded.

  Each processing pass of the S pass, R pass, and C pass has a large contribution to the image quality in this order. Each processing pass is executed in this order, and the context of each coefficient is determined in consideration of information on neighboring coefficients. Hereinafter, a specific processing procedure in each processing path will be described. First, the S pass process will be described.

  When the current coefficient X is insignificant in scanning in each stripe of the bit plane, and at least one of the eight neighboring coefficients adjacent to the current coefficient X is significant, the current coefficient X is It is selected as an S pass processing target. In other cases, it is not an S pass processing target. A coefficient to be decoded in the S pass is a coefficient that is most likely to be significant in the bit plane.

  In the coefficient bit modeling process, for the current coefficient X, Σhi (= h0 + h1) that gives the number of significant coefficients among the two neighboring coefficients h0 and h1 in the horizontal direction, and two neighboring coefficients v0, Σvi (= v1 + v2) that gives the number of significant coefficients in v1, and Σdi (= d1 + d2 + d3 + d4) that gives the number of significant coefficients among the four neighboring coefficients d0, d1, d2, and d3 in the diagonal direction Ask for.

  FIG. 6 is a diagram for explaining a context classification table in the S pass. This context table shows the number of neighborhood coefficients Σhi, Σvi that are significant in the horizontal, vertical, and diagonal directions depending on whether the code block being processed belongs to the LH / LL subband, HL subband, or HH subband. , Σdi and nine kinds of context labels CX identified by labels 0 to 8 are associated with each other. The context estimation unit 22 refers to the context classification table, and determines nine types of coefficients to be processed in the S path according to the combination of the directionality of subband frequency components and the number of neighboring coefficients Σhi, Σvi, Σdi that are significant. And the corresponding context label CX is output.

  In the S pass, when the value of the bit of the current coefficient X is “1”, the significance state of the current coefficient X is changed from “not significant” to “significant”. When the current coefficient X becomes significant, the polarity bit indicating the sign of the current coefficient X is subsequently decoded.

  The context for the polarity bit is determined in the following two steps based on the significance state and the positive and negative polarity values of the two neighboring coefficients v0, v1 in the vertical direction and the two neighboring coefficients h0, h1 in the horizontal direction.

  FIG. 7 is a diagram for explaining a vertical index and a horizontal index table for determining the context of the polarity bit. The context estimation unit 22 first determines whether each of the two vertical neighborhood coefficients v0 and v1 is a positive significant coefficient, a negative significant coefficient, or a nonsignificant coefficient. Referring to the table of FIG. 7, any value of “0”, “1”, “−1” is assigned to the vertical index v. Similarly, any one of “0”, “1”, and “−1” is assigned to the horizontal index h with reference to the table of FIG.

  FIG. 8 is a table in which combinations of horizontal indices h and vertical indices v are associated with context labels CX for five types of polarity bits identified by labels 9 to 13. In the coefficient bit modeling process, referring to the table of FIG. 8, the polarity bit is classified into one of five types of contexts according to the combination of the obtained values of the horizontal index h and the vertical index v, and the corresponding context label CX is output. To do.

  Next, R path processing will be described. When it is determined that the current coefficient X is already significant, the current coefficient X becomes an R path processing target in the bit plane being processed. However, a coefficient that has changed to a significant state in the immediately preceding S pass in the bit plane is not processed. The R-pass decoding process has a role of increasing the accuracy of the coefficients decoded in the S-pass.

  FIG. 9 is a diagram showing a context classification table in the R path. In the R path, in the coefficient bit modeling process, the value of Σhi + Σvi + Σdi indicating the number of significant coefficients is obtained from eight neighboring coefficients with respect to the current coefficient X, and the current coefficient X becomes significant in the bit plane one level higher. In other words, it is determined whether the current R path is the first path after the current coefficient X has changed from a non-significant state to a significant state, and the results are shown in FIG. As described above, the three types of contexts identified by the labels 14 to 16 are classified, and the corresponding context labels CX are output.

  Finally, the C pass process will be described. The C path is used when the current coefficient X does not correspond to either the S path or the R path. In the C pass, run length decoding or decoding processing referring to the values of eight neighboring coefficients as in the S pass is performed. In the C pass, the process proceeds while determining whether to perform run-length decoding.

  FIG. 10 is a diagram illustrating a context classification table in the C path. In the C path, all four consecutive coefficients in the vertical direction are not significant, that is, all the vertical columns in the stripe belong to the C path, and the eight neighboring coefficients for all the four coefficients are also significant. It is determined whether or not the condition that there is no coefficient is. When this condition is satisfied, the decoding process is performed in the run length mode, and the context label CX is 17. If this condition is not met, there is at least one significant coefficient among the four consecutive coefficient values in the vertical direction. Therefore, the position of the first significant coefficient is determined using the UNIFORM context, and the context label at this time CX is 18. The same processing as in the S pass is performed for all remaining coefficients.

  As described above, in the coefficient bit modeling process, the context label CX of each coefficient on the bit plane is determined using any one of the three processing paths. Then, the context label CX is given to the entropy decoding means 3.

FIG. 19 is a diagram illustrating an example of a device configuration according to the present embodiment. An HDD, RAM (inside PC), and CPU (inside PC) are connected via a data bus, and the original code is expanded and color-converted in the following flow. Alternatively, the decompressed data is color-converted.
SP1: Original code data or decompressed data recorded on the HDD is read onto the RAM by a command from the CPU.
SP2: The CPU reads the code on the RAM and performs expansion / color conversion. Alternatively, the CPU reads the decompressed data on the RAM and performs color conversion.
SP3: The CPU writes the data after decompression / color conversion and the data after color conversion in another area on the RAM.
SP4: Data after decompression / color conversion and data after color conversion are recorded on the HDD according to a command from the CPU.

  Here, although the number of bits of the coefficient buffer to be stored by the processing is described above, it will be described briefly. Consider a bit configuration of 9 × 7 wavelet coefficients, which is an input original image of 8 bits at the time of encoding. Since the 9 × 7 wavelet coefficient is a real number, both the integer part and the decimal part need to be stored in the coefficient buffer. The number of bits increased by the wavelet transform is LL1 bit, HL2 bit, LH2 bit, and HH3 bit. Therefore, the maximum number of bits after wavelet transformation is LL9 bits, HL10 bits, LH10 bits, and HH11 bits.

  Next, consider storing coefficients after linear quantization in the same coefficient buffer. Although the number of bits is increased by linear quantization, since the coefficient is converted to an integer for bit plane conversion, it is not necessary to store the decimal part in the coefficient buffer. The number of bits increased by quantization is LL5 bits, HL4 bits, LH4 bits, and HH3 bits. Therefore, the maximum number of bits after quantization is LL14 bits, HL14 bits, LH14 bits, and HH14 bits.

  That is, at the time of decoding, the maximum number of coefficient buffers input before decoding is LL14 bits, HL14 bits, LH14 bits, and HH14 bits. When the original image is 8 bits, a 16-bit coefficient buffer is secured, so that 2 bits of empty bits can be created in the buffer. By using the empty bit, r after the decimal point can be added. A detailed processing method will be described later using a flowchart.

  By using the empty bit, r after the decimal point can be added. A detailed processing method will be described later using a flowchart.

  Here, the flow of the entropy decoding means 3 that sequentially decodes each path in the order of the input paths will be described with reference to the flowchart of FIG.

  FIG. 15 shows a state in which code data is divided into sub-bit planes. This shows a case where the number of sub bit planes is discarded and the number of remaining sub bit planes is 4 when the code data of 03 → 15 → 24. Here, the number of remaining sub-bit planes is the number of remaining sub-bit planes to be decoded by the encoding pass. The coding pass in the code data is sequentially input as 03 → 15 → 24... And processed according to the following flow. For example, 03 includes a C path, C path, 15 includes a C path and S path, 24 includes a C path and an R path, and the encoding paths are sequentially input.

  In this case, the number of remaining sub-bitplanes is four. It is determined whether or not the number of remaining sub-bitplanes of the input code data (here, 03) ≦ 3 (step S1101). If the number of remaining sub-bitplanes ≦ 3 (step S1101, YES), the C path is decoded and the r value is added (step S1102). If the number of remaining sub-bitplanes ≦ 3 is not satisfied (step S1101, NO), the C path decoding process is performed (step S1103). Here, since the number of remaining sub-bitplanes ≦ 3 (the number of remaining sub-bitplanes = 4), the C-pass decoding process is performed. Since the C-pass decoding process was performed, the number of remaining sub-bitplanes was 3.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1104). If the number of remaining sub bit planes> 0 (YES in step S1104), it is determined whether the next code data is an S pass (step S1105). If the number of remaining sub-bitplanes is not greater than 0 (step S1104, NO), the process ends. Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 3), it is determined whether the next code data is an S pass (step S1105). If it is the S pass, it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1106). If the number of remaining sub-bitplanes ≦ 3 (step S1105, YES), the S path is decoded and the r value is added (step S1107). If the number of remaining sub-bitplanes is not ≦ 3, the S-pass decoding process is performed (step S1108). If it is not the S pass (step S1105, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1109). Here, since it is not the S path, the process proceeds to determination of whether the number of remaining sub-bitplanes> 0. Here, the number of remaining sub-bitplanes is 2.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1109). When the number of remaining sub-bitplanes> 0 (step S1109, YES), it is determined whether the next code data is an R path (step S1110). If the number of remaining sub-bitplanes is not greater than 0 (step S1109, NO), the process ends (step S1119). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 2), it is determined whether the next code data is an R path (step S1110). If it is the R path (step S1110, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1111). If the number of remaining sub-bitplanes ≦ 3 (YES in step S1111), the R path decoding process is performed and the r value is added (step S1112). If the number of remaining sub-bitplanes ≦ 3 is not satisfied (step S1111, NO), R path decoding processing is performed (step S1113).

  If it is not an R path (step S1110, NO), the process proceeds to a determination of whether the number of remaining sub-bitplanes> 0 (step S1114). Here, since it is not an R path, the process proceeds to determination of whether the number of remaining sub-bitplanes> 0. Here, the number of remaining sub-bitplanes is 1. Next, it is determined whether or not the number of remaining sub bit planes> 0 (step S1114). When the number of remaining sub-bitplanes> 0 (step S1114, YES), it is determined whether the next code data is a C path (step S1115). If the number of remaining sub-bitplanes is not greater than 0, the process ends (S1119). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 1), it is determined whether the next code data is a C path (step S1115).

  If it is a C path (step S1115, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1116). If the number of remaining sub-bitplanes ≦ 3 (step S1116, YES), the C path decoding process is performed and the r value is added (step S1117). When the number of remaining sub-bitplanes is not 3 (step S1116, NO), the C path decoding process is performed (step S1118). If it is not the C path (step S1115, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1104). Here, since it is the C path, it is determined whether or not the number of remaining sub bit planes ≦ 3 (step S1116). Since the number of remaining sub-bitplanes ≦ 3 (number of remaining sub-bitplanes = 1) (step S1116, YES), the C-pass decoding process is performed and the r value is added (step S1117). Since the C-pass decoding process was performed, the number of remaining sub-bitplanes became zero.

Here, a method of adding r values will be described. The addition value r is arbitrary 0 <r <1. When r is 1/2, r · 2 ^ (number of bit planes−number of decoded bit planes) is added to the least significant bit plane to be processed. Actually, the following values are added.
Addition value = r · 2 ^ (number of bit planes−number of decoded bit planes)
= 1/2 ・ 2 ^ (5-2)
= 4

  When the decoded value is 0, no addition is performed. Here, since the decoded value is 0, no addition is performed.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1104). When the number of remaining sub bit planes> 0 (step S1104, YES), it is determined whether the next code data is an S pass (step S1105). If the number of remaining sub-bitplanes is not greater than 0 (step S1104, NO), the process ends (step S1119). Here, since the number of remaining sub-bitplanes> 0 (the number of remaining sub-bitplanes = 0) is not satisfied, the processing is terminated.

  When the first encoded data (here, 03) is completed, the next encoded data (here, 15) is processed in the same manner as described above. As above, the initial number of remaining sub-bitplanes is four. It is determined whether or not the number of remaining sub-bit planes of the input code data (15 in this case) ≦ 3 (step S1101). Here, since the number of remaining sub-bitplanes ≦ 3 (number of remaining sub-bitplanes = 4) is not satisfied (step S1101, NO), the C path decoding process is performed (step S1103). Since the C-pass decoding process was performed, the number of remaining sub-bitplanes was 3.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1104). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 3) (YES in step S1104), it is determined whether the next code data is an S pass (step S1105). If it is S pass (step S1105, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1106). Here, since the number of remaining sub-bitplanes ≦ 3 (step S1106, YES), the S-pass decoding process is performed and the r value is added (step S1107). Since the S-pass decoding process was performed, the number of remaining sub-bitplanes was 2.

Here, a method of adding r values will be described. The addition value r is arbitrary 0 <r <1. When r is 1/2, r · 2 ^ (number of bit planes−number of decoded bit planes) is added to the least significant bit plane to be processed. Actually, the following values are added.
Addition value = r · 2 ^ (number of bit planes−number of decoded bit planes)
= 1/2 ・ 2 ^ (5-2)
= 4
When the decoded value is 0, no addition is performed.

  Here, before the addition, it is determined whether or not a bit exists in the bit plane below the least significant bit plane. If the bit exists, the added value is added after subtracting the bit. Here, it is 01000, and there is no bit in the bit plane below the least significant bit plane (000), so the subtraction is not performed and the addition is performed as it is. Therefore, it is 01000 before the addition and 01100 after the addition.

  Next, it is determined whether the number of remaining sub bit planes> 0 (S1109). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 2) (YES in step S1109), it is determined whether the next code data is an R path (step S1110). Here, since it is not an R path, the process proceeds to determination of whether the number of remaining sub-bitplanes> 0. Here, the number of remaining sub-bitplanes is 1.

  Next, it is determined whether or not the number of remaining sub bit planes> 0 (step S1114). Here, since the number of remaining sub-bit planes> 0 (number of remaining sub-bit planes = 1) (step S1114, YES), it is determined whether the next code data is a C path (step S1115). Here, since it is not the C path, the process proceeds to a determination as to whether the number of remaining sub bit planes> 0 (step S1104). Here, the number of remaining sub-bitplanes is zero. Next, it is determined whether the number of remaining sub bit planes> 0 (step S1104). Here, since the number of remaining sub-bitplanes> 0 (the number of remaining sub-bitplanes = 0), the process ends (step S1119).

  When the second encoded data (here 15) is completed, the next encoded data (here 24) is processed in the same manner as described above. Processing is performed until all encoded data to be input is completed.

Here, the case where the code data as shown in FIG. 13 is divided into sub-bit planes will be described. This shows the case where the number of sub-bit planes is discarded by 2 and the number of remaining sub-bit planes is 8 when the code data of 01 → 07 → 12. Here, the number of remaining sub-bit planes is the number of remaining sub-bit planes to be decoded by the encoding pass. The encoding pass in the encoded data is sequentially input as 01 → 07 → 12... And processed according to the following flow. For example, 01 is a C path, C path, S path, S path, 07 is a C path, S path, R path, 12 is a C path, R path, R path, etc. The encoding pass is input sequentially.

In this case, the number of remaining sub bit planes is eight.
It is determined whether or not the number of remaining sub-bit planes of the input code data (here, 03) ≦ 3 (S1101). If the number of remaining sub-bitplanes ≦ 3 (step S1101, YES), the C path is decoded and the r value is added (step S1102). If the number of remaining sub-bitplanes ≦ 3 is not satisfied (step S1101, NO), the C path decoding process is performed (step S1103). Here, since the number of remaining sub-bitplanes ≦ 3 (the number of remaining sub-bitplanes = 8), the C path decoding process is performed. Since the C-pass decoding process was performed, the number of remaining sub-bitplanes was 7.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1104). If the number of remaining sub bit planes> 0 (step S1105, YES), it is determined whether the next code data is an S pass (step S1105). If the number of remaining sub-bitplanes is not greater than 0, the process ends (step S1119). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 7), it is determined whether the next code data is an S pass (step S1105). If it is S pass (step S1105, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1106).

  If the number of remaining sub-bitplanes ≦ 3 (step S1106, YES), the S path is decoded and the r value is added (step S1107). If the number of remaining sub-bitplanes is not 3 (step S1106, NO), the S path decoding process is performed (step S1108). If it is not the S pass (step S1105, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1109). Here, since it is not the S path, the process proceeds to determination of whether the number of remaining sub-bitplanes> 0. Here, the number of remaining sub-bitplanes is 6.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1109). When the number of remaining sub-bitplanes> 0 (step S1109, YES), it is determined whether the next code data is an R path (step S1110). If the number of remaining sub-bitplanes is not greater than 0, the process ends (step S1119). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 6) (YES in step S1109), it is determined whether the next code data is an R path (step S1110). If it is the R path (step S1110, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1111).

  If the number of remaining sub-bitplanes ≦ 3 (YES in step S1111), the R path decoding process is performed and the r value is added (step S1112). If the number of remaining sub-bitplanes ≦ 3 is not satisfied (step S1111, NO), R path decoding processing is performed (step S1113). If it is not an R path (step S1110, NO), the process proceeds to a determination of whether the number of remaining sub-bitplanes> 0 (step S1114). Here, since it is not an R path, the process proceeds to determination of whether the number of remaining sub-bitplanes> 0. Here, the number of remaining sub-bitplanes is 5.

Next, it is determined whether or not the number of remaining sub bit planes> 0 (step S1114). When the number of remaining sub-bitplanes> 0 (step S1114, YES), it is determined whether the next code data is a C path (step S1115). If the number of remaining sub-bitplanes is not greater than 0 (step S1114, NO), the process ends (step S1119). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 5) (YES in step S1114), it is determined whether the next code data is a C path (step S1115).
If it is a C path (step S1115, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1116).

  If the number of remaining sub-bitplanes ≦ 3 (step S1116, YES), the C path decoding process is performed and the r value is added (step S1117). When the number of remaining sub-bitplanes is not 3 (step S1116, NO), the C path decoding process is performed (step S1118). If it is not the C path (step S1115, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1104).

  Here, since it is the C path, it is determined whether or not the number of remaining sub bit planes ≦ 3 (step S1116). Since the number of remaining sub-bitplanes ≦ 3 (number of remaining sub-bitplanes = 5) (YES in step S1116), the C path decoding process is performed (step S1117). Since the C-pass decoding process was performed, the number of remaining sub-bitplanes was 4.

  The above process is repeated until the number of remaining sub bit planes becomes zero. A case where the number of remaining sub-bitplanes is 1 will be described by repeating the S, R, and C passes.

  It is determined whether the number of remaining sub bit planes> 0 (step S1104). If the number of remaining sub bit planes> 0 (YES in step S1104), it is determined whether the next code data is an S pass (step S1105). If the number of remaining sub-bitplanes is not greater than 0, the process ends (step S1119). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 1) (YES in step S1104), it is determined whether the next code data is an S pass (step S1105). If it is S pass (step S1105, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1106).

  If the number of remaining sub-bitplanes ≦ 3 (step S1106, YES), the S path is decoded and the r value is added (step S1107). If the number of remaining sub-bitplanes is not ≦ 3, the S-pass decoding process is performed (step S1108). If it is not the S pass (step S1105, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1109). Here, since it is the S path (step S1105, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1106). Since the number of remaining sub-bit planes ≦ 3 (the number of remaining sub-bit planes = 1), an S-pass decoding process is performed and an r value is added (step S1107). Since the S-pass decoding process was performed, the number of remaining sub-bitplanes was zero.

Here, a method when the r value is added will be described. The addition value r is arbitrary 0 <r <1. When r is 1/2, r · 2 ^ (number of bit planes−number of decoded bit planes) in the least significant bit plane to be processed
Is added. Actually, the following values are added.
Addition value = r · 2 ^ (number of bit planes−number of decoded bit planes)
= 1/2 ・ 2 ^ (4-4)
= 0
When the decoded value is 0, no addition is performed.

  Here, the flow of flag insertion into empty bits will be described with reference to the flowchart of FIG. The condition whether or not the addition processing in step S1701 is necessary is when bit plane = 0 and addition value = 0. The flowchart of FIG. 17 is omitted in the above description, but is performed during the process of performing the r value addition process in each of the steps S1102, S1107, S1112, and S1117 of FIG.

  Since bit plane = 0 and addition value = 0 (step S1701, YES), the process proceeds to the process of inserting a flag into an empty bit (step S1702) for the addition process. Therefore, a flag is set for the empty bit of the code data. For example, 1 is put in the upper 2 bits of the code data. Next, it is determined whether or not the number of remaining sub bit planes> 0 (step S1114). Here, since the number of remaining sub-bitplanes> 0 (the number of remaining sub-bitplanes = 0), the process ends (step S1119). When the first encoded data (here 01) is completed, the next encoded data (here 07) is processed in the same manner as described above.

  The flow of r value addition processing using an empty bit flag will be described with reference to the flowchart of FIG. In the process behind the above process, the presence or absence of a flag is determined (step S1801). If the flag is present (YES in step S1801), the bit plane is 0 or less, that is, the r value of the decimal point is added (step S1802). be able to. Here, since the r value is ½, ½ should be added.

  Here, the reason why the number of remaining sub bit planes ≦ 3 will be described. Each encoded data exists in one of the SRC paths. Therefore, the addition process is performed on the upper three sub-bit planes from the least significant sub-bit plane, thereby minimizing the addition process.

  Therefore, it is possible to accurately determine a bit plane that is a truncation point, and to suppress degradation in image quality in decoding processing by adding an arbitrary value r to lower-order bit data than the truncation point using the bit plane. Bit planes that are likely to be added are estimated based on the number of planes, and addition processing can be performed only when it is at least necessary, which is a high-speed processing method.

<Configuration of Second Embodiment>
Here, the flow of the entropy decoding means 3 that sequentially decodes each path in the order of the input paths will be described with reference to the flowchart of FIG.

  FIG. 14 shows a state in which code data is divided into sub-bit planes. This shows the case where the number of sub-bit planes is discarded by 2 and the number of remaining sub-bit planes is 8 when the code data of 01 → 07 → 12. Here, the number of remaining sub-bit planes is the number of remaining sub-bit planes to be decoded by the encoding pass. The encoding pass in the encoded data is sequentially input as 01 → 07 → 12... And processed according to the following flow. For example, 01 is C path, C path, S path, S path, 07 is C path, S path, R path, R path, 12 is C path, R path, R path, R path, etc. Paths are included, and encoding paths are sequentially input.

  In this case, the number of remaining sub bit planes is nine. It is determined whether or not the number of remaining sub-bit planes of the input code data (here, 03) ≦ 3 (step S1201). When the number of remaining sub-bitplanes ≦ 3 (step S1201, YES), the C path decoding process is performed and the r value is added (step S1202). When the number of remaining sub-bitplanes is not 3 (step S1201, NO), C path decoding processing is performed (step S1203). Here, since the number of remaining sub-bitplanes ≦ 3 (the number of remaining sub-bitplanes = 9) is not satisfied (step S1201, NO), the C path decoding process is performed. Since the C-pass decoding process was performed, the number of remaining sub-bitplanes was 8.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1204). If the number of remaining sub bit planes> 0 (step S1204, YES), it is determined whether the next code data is an S pass (step S1205). If the number of remaining sub-bitplanes is not greater than 0 (step S1204, NO), the process ends (S1219). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 8), it is determined whether the next code data is an S pass (step S1205). If it is S pass (step S1205, YES), it is determined whether the number of remaining sub-bitplanes ≦ 4 (step S1206).

  If the number of remaining sub-bitplanes ≦ 4 (YES in step S1206), the S path is decoded and the r value is added (step S1207). If the number of remaining sub-bitplanes is not 4 (step S1206, NO), the S path decoding process is performed (step S1208). If it is not the S path (step S1205, NO), the process proceeds to a determination of whether the number of remaining sub-bitplanes> 0 (step S1209). Here, since it is not the S path, the process proceeds to determination of whether the number of remaining sub bit planes> 0 (step S1209). Here, the number of remaining sub-bitplanes is 7.

  Next, it is determined whether or not the number of remaining sub bit planes> 0 (step S1209). When the number of remaining sub-bitplanes> 0 (step S1209, YES), it is determined whether the next code data is an R path (step S1210). If the number of remaining sub bit planes is not greater than 0 (step S1209, NO), the process ends (step S1219). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 7), it is determined whether the next code data is an R path (step S1210). If it is an R path (step S1210, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1211).

  If the number of remaining sub-bitplanes ≦ 3 (step S1211, YES), the R path decoding process is performed and the r value is added (step S1212). When the number of remaining sub-bitplanes is not 3 (step S1211, NO), R path decoding processing is performed (step S1213). If it is not the R path (step S1210, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1214). Here, since it is not an R path, the process proceeds to determination of whether the number of remaining sub bit planes> 0 (step S1214). Here, the number of remaining sub-bitplanes is 6.

  Next, it is determined whether the number of remaining sub bit planes> 0 (step S1214). When the number of remaining sub bit planes> 0 (step S1214, YES), it is determined whether the next code data is a C path (step S1215). If the number of remaining sub-bitplanes is not greater than 0 (step S1214, NO), the process ends (step S1219). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 6), it is determined whether the next code data is a C path (step S1215). If it is a C path (step S1215, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1216).

  When the number of remaining sub bit planes ≦ 3 (step S1216, YES), the C path decoding process is performed, and the r value is added (step S1217). When the number of remaining sub-bitplanes is not 3 (step S1216, NO), the C path decoding process is performed (step S1218). If it is not the C path (step S1215, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1204). Here, since it is a C path, it is determined whether the number of remaining sub bit planes ≦ 3 (step S1216). Since the number of remaining sub-bitplanes ≦ 3 (number of remaining sub-bitplanes = 6) (step S1216, NO), the C path decoding process is performed. Since the C-pass decoding process was performed, the number of remaining sub-bitplanes was 5.

  The above process is repeated until the number of remaining sub bit planes becomes zero. A case where the number of remaining sub-bitplanes is 2 will be described by repeating the S, R, and C passes.

  It is determined whether the number of remaining sub bit planes> 0 (step S1204). If the number of remaining sub bit planes> 0 (step S1204, YES), it is determined whether the next code data is an S pass (step S1205). If the number of remaining sub-bitplanes is not greater than 0 (step S1204, NO), the process ends (step S1219). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 2), it is determined whether the next code data is an S pass (step S1205). If it is S pass (step S1205, YES), it is determined whether the number of remaining sub-bitplanes ≦ 4 (step S1206).

  If the number of remaining sub-bitplanes ≦ 4 (YES in step S1206), the S path is decoded and the r value is added (step S1207). If the number of remaining sub-bitplanes is not 4 (step S1206, NO), the S path decoding process is performed (step S1208). If it is not the S path (step S1205, NO), the process proceeds to a determination of whether the number of remaining sub-bitplanes> 0 (step S1209). Here, since it is the S pass, it is determined whether the number of remaining sub bit planes ≦ 4 (step S1206). Since the number of remaining sub-bitplanes ≦ 4 (number of remaining sub-bitplanes = 2), S-path decoding processing is performed and r-value addition is performed (step S1207). Since the S-pass decoding process was performed, the number of remaining sub-bitplanes was 1.

Here, a method when the r value is added will be described. The addition value r is arbitrary 0 <r <1. When r is 1/2, r · 2 ^ (number of bit planes−number of decoded bit planes) is added to the least significant bit plane to be processed. Actually, the following values are added.
Addition value = r · 2 ^ (number of bit planes−number of decoded bit planes)
= 1/2 ・ 2 ^ (4-4)
= 0
When the decoded value is 0, no addition is performed.

  Here, the flow of flag insertion into empty bits will be described with reference to the flowchart of FIG. Here, the conditions for whether or not the addition processing in step S1701 is necessary are bit plane = 0 and addition value = 0. The flowchart of FIG. 17 is omitted in the above description, but is performed during the process of performing the r value addition process in each of the steps S1102, S1107, S1112, and S1117 of FIG.

  Since the bit plane = 0 and the addition value = 0, the process proceeds to the flag insertion (step S1702) processing into the empty bit for the addition processing (step S1701). Therefore, a flag is set for the empty bit of the code data. For example, 1 is put in the upper 2 bits of the code data.

  Next, it is determined whether or not the number of remaining sub bit planes> 0 (step S1209). If the number of remaining sub-bitplanes> 0 (step S1209, YES), it is determined whether the next code data is an R path (step S1210). If the number of remaining sub-bitplanes is not greater than 0 (step S1209, NO), the process ends (step S1219). Here, since the number of remaining sub bit planes> 0 (the number of remaining sub bit planes = 1), it is determined whether the next code data is an R path (step S1210).

  If it is the R path (step S1210, YES), it is determined whether the number of remaining sub-bitplanes ≦ 3 (step S1206). If the number of remaining sub-bitplanes ≦ 3 (step S1210, YES), the R path is decoded and the r value is added (step S1212). If the number of remaining sub-bitplanes is not 3 (step S1211, NO), the R path decoding process is performed (step S1213). If it is not the R path (step S1210, NO), the process proceeds to determination of whether the number of remaining sub-bitplanes> 0 (step S1214). Here, since it is not an R path, the process proceeds to determination of whether the number of remaining sub-bitplanes> 0. Here, the number of remaining sub-bitplanes is zero.

  Next, since the number of remaining sub-bit planes> 0 (number of remaining sub-bit planes = 0) (step S1214, NO), the process ends (step S1219). When the first encoded data (here 01) is completed, the next encoded data (here 07) is processed in the same manner as described above.

  The flow of r value addition processing using an empty bit flag will be described with reference to the flowchart of FIG. In the process after the above process, the presence / absence of an empty bit flag is determined (S1801). If there is a flag (YES in step S1801), the bit plane is 0 or less, that is, the r value of the decimal point is added (step S1802). )can do. Here, since the r value is ½, ½ should be added.

  Here, the reason why the number of remaining sub-bit planes ≦ 4 in step S1206 will be described. Each encoded data exists in one of the SRC paths. Therefore, the addition process is performed on the upper three sub-bit planes from the least significant sub-bit plane, thereby minimizing the addition process. However, each encoded data may not exist in any of the SRC paths depending on the truncation point of the sub bit plane. This is a case where the data is truncated by the S pass as shown in FIG. In the case of the encoded data 15, there is no path in any of the upper three paths (in order of RCS) from the least significant sub-bit plane. This is because, in the case of the S path, the next path that comes out is the S path or the R path, so the next path that comes out is the R path, such as the upper fourth S path from the least significant subbit plane. If there is, the addition process is not performed. Therefore, by performing addition processing on the upper four sub-bit planes (in order of SRCS) from the least significant sub-bit plane, the minimum addition processing can be performed.

  The bit plane that becomes the truncation point is accurately determined, and an arbitrary r addition value to the lower bit data than the truncation point using the bit plane suppresses the image quality degradation of the decoding process, and the number of remaining sub bit planes Thus, it is possible to estimate a bit plane that is likely to be added, and to perform addition processing only when it is at least necessary, which is a high-speed processing method.

  Further, according to the above embodiment, a flag is necessary to perform addition of r that accurately determines a bit plane that is a truncation point, and to perform addition even when r is a decimal point or less. In addition, by effectively using the empty bits for the flag, it is possible to suppress image quality deterioration, perform high-speed processing, and effectively use the memory.

  Further, the entropy encoding means of the present invention determines whether or not to perform the reconstruction value addition process after the encoding pass process in the encoding pass process in the entropy decoding process. A flag may be set to the vacant bit of the determined data.

  In the present invention, whether to perform reconstruction value addition processing may be determined based on the number of remaining sub-bitplanes.

  In the present invention, the magnitude refining path and the cleanup path of the entropy encoding means may be set to 3 and the significant propagation path may be set to 4.

  Further, the entropy encoding means of the present invention determines the presence or absence of a flag in a vacant bit after processing an encoding pass, and if a vacant bit has a flag, a predetermined addition value other than a predetermined zero is determined. It may be added to the decoded data having the value of.

Also, the entropy encoding process of the present invention, in the process of the encoding pass of the entropy decoding process,
It may be determined whether or not the reconstruction value addition process after the encoding pass process is performed, and in the case of performing the addition process, a flag may be set to the vacant bit of the determined data.

  In the present invention, whether to perform reconstruction value addition processing may be determined based on the number of remaining sub-bitplanes.

  In the present invention, the magnitude relocation path and cleanup path of the entropy encoding process may be set to 3, and the significant propagation path may be set to 4.

  Further, according to the present invention, the entropy encoding step determines whether or not there is a flag in a vacant bit after processing the encoding pass. If the vacant bit has a flag, a predetermined addition value is set to a predetermined 0. It may be added to decoded data having a value other than.

Further, the entropy encoding process of the present invention is performed in the encoding pass process in the entropy decoding process.
It may be determined whether or not the reconstruction value addition process after the encoding pass process is performed, and in the case of performing the addition process, a flag may be set to the vacant bit of the determined data.

  In the present invention, the determination as to whether or not to perform reconstruction value addition processing may be based on the number of remaining sub-bitplanes.

  In the present invention, the magnitude refining path and the cleanup path of the entropy encoding process may be set to 3 and the significant propagation path may be set to 4.

  Also, the entropy encoding process of the present invention determines the presence or absence of a flag in a vacant bit after processing an encoding pass, and if a vacant bit has a flag, a predetermined addition value other than a predetermined zero is determined. It may be added to the decoded data having the value of.

  The present invention can be applied to uses such as an image encoding / decoding device, an image processing device, and an image forming device.

DESCRIPTION OF SYMBOLS 1 Code data input means 2 Header input means 3 Entropy decoding means 4 Scalar inverse quantization means 5 Inverse wavelet transform means 6 Color conversion means 7 DC level shift means 8 Image data output means

JP 2003-189107 A JP 2006-33179 A

Jiro Katto, Yasuhiko Yasuda, Characteristic evaluation of subband coding and optimization of its filter coefficients, IEICE Technical Report, May. 1991. J. et al. Katto and Y.C. Yasuda, Performance Evaluation of Subband Coding and Optimization of It's Filter Coefficients, SPIE VCIP '91, Nov. 1991. Kuniaki Komatsu, Akira Sezaki, Optimal quantization step size for integer lossless transform coefficients, IEICE Transactions A, vol. J85-A, no. 11, pp. 1182-1189, 2002.11. Davi Taubman, Michael Marcellin, JPEG2000 Image Compression Fundamentals, Standards and Practice, Kluwer Academic Publishers.

Claims (17)

In an apparatus for processing and encoding / decoding the image data by JPEG2000, which is an encoding method for encoding image data,
Fixed decimal arithmetic means for performing fixed decimal arithmetic processing for arithmetic processing of the image data;
9 × 7 filter means used for arithmetic processing for performing wavelet transform of the image data;
A scalar quantization means for performing a scalar quantization process which is an arithmetic process for quantizing the image data;
Entropy encoding means for performing entropy encoding processing that is arithmetic processing for encoding / decoding the image data;
Data buffer means for storing data subjected to arithmetic processing for performing arithmetic processing of the image data;
Have
The data buffer means uses a fixed number of bits for storing the processed data, and uses empty bits in the data buffer for decoding processing. The entropy encoding means, in the process of the encoding pass of the entropy decoding process,
The determination whether or not to perform a reconstruction value addition process after the process of the encoding pass is performed, and when the addition process is performed, a flag is set in an empty bit of the determined data. The decoding device according to 1.   The decoding apparatus according to claim 1 or 2, wherein whether or not to perform the reconstruction value addition processing is determined based on the number of remaining sub-bitplanes.   The decoding apparatus according to any one of claims 1 to 3, wherein a magnitude refinement path and a clean-up path of the entropy encoding means are set to 3, and a significance propagation path is set to 4. .   The entropy encoding means determines the presence or absence of a flag in a vacant bit after processing the encoding pass, and if a vacant bit has a flag, sets a predetermined addition value to a predetermined value other than 0. The decoding apparatus according to any one of claims 1 to 4, wherein the decoding apparatus adds the decoded data.   An image processing apparatus comprising the encoding / decoding apparatus according to claim 1. In a method of processing the image data by JPEG2000, which is an encoding method for encoding the image data, and encoding / decoding the image data,
A fixed decimal number calculation step for performing a fixed decimal number calculation process for calculating the image data;
A 9 × 7 filter step used for arithmetic processing for performing wavelet transform of the image data;
A scalar quantization step for performing a scalar quantization process which is an arithmetic process for quantizing the image data;
An entropy encoding step for performing an entropy encoding process that is an arithmetic process for encoding / decoding the image data;
A data buffer step for storing data processed for performing the arithmetic processing of the image data, using a fixed number of bits, and using empty bits of the data buffer for decoding processing;
A coding / decoding method characterized by comprising: In the entropy encoding process, in the process of the encoding pass of the entropy decoding process,
The determination whether or not to perform a reconstruction value addition process after the process of the encoding pass is performed, and when the addition process is performed, a flag is set in an empty bit of the determined data. 8. The decoding method according to 7.   The decoding method according to claim 7 or 8, wherein whether or not to perform the reconstruction value addition processing is determined based on the number of remaining sub-bitplanes.   The decoding method according to any one of claims 7 to 9, wherein a magnitude refinement path and a cleanup path in the entropy encoding step are set to 3, and a significant propagation path is set to 4. .   The entropy encoding step determines the presence or absence of a flag in a vacant bit after processing the encoding pass. If the vacant bit has a flag, a predetermined addition value is set to a predetermined value other than 0. The decoding method according to any one of claims 7 to 10, wherein the decoding data is added to the decoded data. In a program for processing and encoding / decoding the image data by JPEG2000, which is an encoding method for encoding image data,
Fixed decimal arithmetic processing for performing fixed decimal arithmetic processing for arithmetic processing of the image data;
9 × 7 filter processing used for arithmetic processing for performing wavelet transform of the image data;
A scalar quantization process for performing a scalar quantization process which is an arithmetic process for quantizing the image data;
Entropy encoding processing for performing entropy encoding processing, which is arithmetic processing for encoding and decoding the image data;
A data buffer process for using a fixed number of bits for storing the data processed to perform the calculation process of the image data, and for using empty bits of the data buffer for the decoding process;
A program that causes a computer to execute. The entropy encoding process is an encoding pass process in the entropy decoding process.
The determination whether or not to perform a reconstruction value addition process after the process of the encoding pass is performed, and when the addition process is performed, a flag is set in an empty bit of the determined data. 12. The program according to 12.   The program according to claim 12 or 13, wherein whether or not to perform the reconstruction value addition processing is determined based on the number of remaining sub-bitplanes.   The program according to any one of claims 12 to 14, wherein a magnitude refinement path and a cleanup path of the entropy encoding process are set to 3, and a significance propagation path is set to 4.   The entropy encoding process determines the presence / absence of a flag in a vacant bit after processing an encoding pass, and if a vacant bit has a flag, a predetermined addition value is set to a predetermined non-zero value. The program according to any one of claims 12 to 15, wherein the program is added to the decoded data.   A computer-readable recording medium having recorded thereon a program for executing the method according to any one of claims 7 to 11.

Images