JP5200792B2 - Image compression apparatus and method, and program - Google Patents

Description

  The present invention relates to an image compression apparatus and image compression method for compressing and encoding image data in conformity with JPEG2000 and the like, and a program therefor. The present invention relates to a still image or video compression / decompression technique, and in particular, a digital still camera, a recording part of a digital copier, a surveillance camera, a digital video storage, a digital camera (moving image), a satellite communication device, a medical device, and the Internet. It can be applied to image processing in a browser or the like.

  In recent years, JEPPG2000 is known as a compression encoding method suitable for handling high-definition images. In the JPEG2000 encoding process, after image data is converted into Y, Cb, and Cr color component data, two-dimensional discrete wavelet transform is performed on each data as frequency analysis.

  JPEG2000 is a format related to compression / decompression of image data. FIG. 9 is a functional block diagram for explaining the basics of the JPEG2000 algorithm. This functional block diagram shows algorithms corresponding to various functions to be executed by a computer in accordance with an image processing computer program for image compression / decompression in JPEG2000 format.

As shown in FIG. 9, the JPEG2000 algorithm includes a color space transformation / inverse transformation unit 111, a two-dimensional wavelet transformation / inverse transformation unit 112, a quantization / inverse quantization unit 113, an entropy encoding / decoding unit 114, and tag processing. It is comprised by the part 115 (for example, refer nonpatent literature 1).

  Wavelet coefficient data (for example, 16-bit data) obtained by wavelet transform is converted into subbands (for example, in the case of level 3 wavelet transform, 3LL, 3HL, 3LH, 3HH, 2HL, 2LH, 2HH, 1HL, 1LH, 1HH). ) Is divided into bit planes as processing units, and the arithmetic coding is performed by scanning the data of each bit plane for each subband in order from the top by three methods. The above three methods are called “significant propagation pass”, “magnitude refinement pass”, and “cleanup pass”. .

  The compression of the code data is performed by uniformly deleting (truncating) the code data of the coding pass of all the code blocks of each subband obtained by the scan by the above three methods in order from the least significant bit plane side. Here, the deletion of the code data means that the value of the bit data to be deleted is replaced with 0 (invalid data). Refer to Non-Patent Document 1 for details of JPEG2000 encoding processing.

  Further, as conventional documents related to image compression technology, there are, for example, Patent Document 2 and Patent Document 3 in addition to Patent Document 1 described above. The outline of these conventional documents will be briefly described below.

  The one disclosed in Patent Document 1 includes a compression unit that executes compression processing of an image in JPEG 2000 format, and a decompression unit that executes decompression processing of an image compressed in a format other than JPEG 2000, and is stored in a storage area. When it is determined that the image file is an uncompressed file, it is determined that the image file is an uncompressed file or a compressed file compressed in a format other than JPEG2000, and the image file is determined to be an uncompressed file. If the compressed file is determined to be a compressed file compressed in a format other than JPEG2000, it is decompressed by the decompressing means and then compressed by the compressing means, and from the compressed image to the designated region of interest. The corresponding block is extracted and output.

  The one disclosed in Patent Document 2 is generated for each processing unit by an encoding unit that generates coefficient data by encoding coefficient data obtained by frequency analysis of image data for each processing unit, and the encoding unit. A data reduction unit for reducing the amount of code data by deleting the code data, and dividing the coefficient data of each processing unit into coefficient data of the second processing unit, and increasing the value of the coefficient data of the second processing unit Accordingly, a data processing unit for increasing the amount of code data reduction in the data reduction unit by the second processing unit is provided.

  In addition, what is disclosed in Patent Document 3 is an encoding unit that generates code data by encoding coefficient data obtained by frequency analysis of image data in units of processing, and coefficients of each processing unit among the code data. Data that determines the contents to be deleted from the code data corresponding to the data on the lower bit side of the data is data No. Code data when the reduction amount of code data gradually increases or decreases and the quality of the reproduced image gradually deteriorates or improves, and code data when code data is deleted according to the data read from the table The data No. of one data that can be regarded as the target value is calculated. And a rate control unit for specifying the above.

  FIG. 10 is a diagram illustrating the configuration of the image compression apparatus disclosed in Patent Documents 2 and 3. As shown in FIG. 10, the image compression apparatus 100 includes a wavelet transform unit 10, an arithmetic coding unit 20, a packet header generation processing unit 30, a memory controller 40, and a DRAM (Dynamic Random Access) 50. The memory controller 40 is a so-called arbiter circuit, and each DMA (Direct Memory Access) 13, 21, 25, 31, 33, 35 provided in the wavelet transform unit 10, the arithmetic code processing unit 20 and the packet header generation processing unit 30 The arbitration of the data access right to the 37 DRAM 50 is performed.

  The wavelet transform unit 10 transforms image data into 16-bit wavelet coefficients. The color conversion circuit 11 converts input image data into Y, Cb, and Cr color components and outputs them. The wavelet transform circuit 12 performs a two-dimensional discrete wavelet transform on the data of each component after color conversion. The DMA 13 stores the generated wavelet coefficient at a predetermined address of the DRAM 50.

  The arithmetic encoding unit 20 performs arithmetic encoding processing on the wavelet coefficients stored in the DRAM 50 in the wavelet transform unit 10 and writes the code data for each subband coding pass to the DRAM 50, and at the same time, The mask amount (number of coding passes) for each code block is specified, and the specified mask amount is written in the memory A or the memory B.

  Further, from the above-described mask amount and code amount for each coding pass of each code block, a code amount to be reduced when deleting each subband coding pass one by one based on the truncation data is specified, and the memory C or Write to memory D.

  The DMA 21 reads out the wavelet coefficient stored at a predetermined address of the DRAM 50 in units of subbands. The read 16-bit wavelet coefficients are entropy quantized by the quantizing circuit 22 and then input to the bit plane dividing circuit 23 to be divided into 16 bit planes.

  The arithmetic encoding circuit 24 performs arithmetic encoding by scanning the data of each bit plane for each subband in order from the upper bits by three methods (referred to as coding passes). The above three methods are called “significant propagation pass”, “magnitude refinement pass”, and “cleanup pass”. Code data output from the arithmetic coding unit 24 is composed of code data of 15 × 3 + 1 = 46 coding passes, and is written into the DRAM 50 via the DMA 25.

  The wavelet coefficients read from the DRAM 50 by the DMA 21 are input not only to the quantization circuit 22 but also to the average value calculation circuit 26. The average value calculation circuit 26 calculates and outputs the average value of the wavelet coefficients of the effective pixels of each code block. The masking coefficient calculation circuit 27 adds 0, 1 or 2 additional masking coefficients according to the average value of the wavelet coefficients for each code block sequentially output from the average value calculation circuit 26, that is, the mask amount. (Number of coding passes from which code data is deleted) is specified and output. The memory A and the memory B are memories that are alternately enabled in units of frame images to be processed, and record the mask amount for each code block of each subband output from the masking coefficient calculation circuit 27.

  The code data output from the arithmetic encoding circuit 24 is also output to the code amount calculation circuit 28. The code amount calculation circuit 28 counts the code amount for each coding pass corresponding to the bit plane of each code block, and outputs the coefficient value to the data processing circuit 29. Further, the mask amount (the number of coding passes) of each code block written in the memory A or memory B is input to the data processing circuit 29. When deleting the code data of the coding pass of each subband one by one from the least significant bit plane side, the data processing circuit 29 further obtains a reduction amount of the code data considering the mask amount, and obtains the obtained code data Is written in the memory C or the memory D.

  The packet header generation processing unit 30 is a case where the code data of the coding pass of each subband obtained by the arithmetic encoding unit 20 is deleted one by one from the least significant bit plane side, and further, for each code block Based on the reduction amount of the code data when the mask amount is taken into consideration, the data No. of truncation data suitable for deleting the code data by the desired amount is deleted. And the identified data No. A packet header of code data obtained based on the truncation data is generated, and a bit stream is formed and output. First, the rate control circuit 32 first sets the data number. 128 truncation data is read from the DRAM 50 via the DMA 31, and the total reduction amount of code data specified for each code block is calculated for all subbands according to the content of the read truncation data. Make a comparison.

  Here, if the target reduction amount is not reached, a larger data No. The truncation data is read out, and the total data reduction amount in all subbands is obtained according to the content of the data. In the opposite case, that is, when the total value of the reduction amount of the code data is larger than the target reduction amount, the data No. The truncation data is read out, and the reduction amount of the code data is calculated again. The data No. 1 with the code data reduction amount becomes a value that can be regarded as the target reduction amount. This data No. is identified. Is output to the packet information generation circuit 34 in the subsequent stage.

  The rate control circuit 32 is roughly divided into an address generation circuit, a code amount calculation circuit, and a data number. And a switching circuit. The address generation circuit and the code amount calculation circuit have the designated data No. The truncation data is read from the truncation table, and functions as an arithmetic unit that calculates a variable of the code data when deleting the lower bit data of the code data of each processing unit according to the content of the read truncation data. The address generation circuit generates an address for reading out the code amount data of the bit plane to be added or subtracted based on the content of the truncation data input via the DMA 31, and outputs the address to the memory C and the memory D.

  Of the memory C and the memory D, the data of the frame image being processed is stored, and the memory in the enabled state calculates the code amount by calculating the data representing the reduction amount of the code data stored at the specified address. Output to the circuit. The code amount calculation circuit obtains the sum of all the subbands of the data reduction amount sent from the memory C or the memory D, compares the obtained total value with the target reduction amount, and outputs a signal representing the comparison result. No. data No. Output to the switching circuit. Data No. The switching circuit has different data No. based on the comparison result signal output from the code amount calculation circuit. Is requested to the DRAM 50 via the DMA 33. Further, the last data No. 1 in which the code data reduction amount can be regarded as the target reduction amount. Is output to the packet information generation circuit 34.

  The packet header generation circuit 36 outputs the code pass number and code amount data of the code block of each subband output from the packet information generation circuit 34, the number of zero bit planes, and the code data read from the DRAM 5 from the DMA 37. A packet header is generated from the data such as and the like and output to the code forming circuit 38. The code formation circuit 38 forms a bit stream from the data output from the packet header generation circuit 36 and outputs the bit stream to the outside as code data in which the code data has been reduced by the target reduction amount. The above is the outline of the image compression apparatus in Patent Documents 1 and 2.

  JPEG2000 has a function that can improve the image quality of only a “selected portion of an image” called ROI (Region of Interest). In order to realize the ROI function, it is necessary to perform processing corresponding to the encoding. In the basic method of JPEG 2000, a method of weighting the coefficient of the region of interest (specifically, a power of 2 is weighted by bit shift) is adopted.

  FIG. 11 is a diagram showing the flow of JPEG2000 encoding processing according to the prior art. As shown in FIG. 11, the image is wavelet transformed (step S501), quantized (step S502), MQ (arithmetic) coded (step S503), code-formed (step S504), and compressed image data is obtained. (Step S505). Usually, in order to realize the ROI function, it is necessary to perform processing corresponding to the encoding.

  FIG. 12 is a diagram showing a flow including ROI processing performed at the time of encoding. As shown in FIG. 12, the image is wavelet transformed (step S601), bit-shifted (step S602), quantized (step S603), MQ (arithmetic) coded (step S604), and code-formed (step S605). ), Compressed image data was obtained (step S606). Usually, in order to realize the ROI function, it is necessary to perform processing corresponding to the encoding. The processing can be thought of as a method of encoding the region of interest with quantization steps that are finer than other parts and a method of weighting the coefficients of the region of interest (specifically, bit shifting) (Refer to step S602 in FIG. 12).

  The JPEG2000 basic method shown in FIG. 12 employs the latter method. The shift amount of the bit shift is the maximum number of bits of the coefficient that is not the ROI part, and from this point, it is called the Maxshift method. The coefficients in the ROI area need to be returned to the same range as the coefficients in other areas (background areas) at the time of decoding while being shifted up by this bit shift processing.

  FIG. 13 is a diagram showing the flow of ROI processing at the time of decoding, and basically the processing is performed in the reverse order of FIG. That is, code analysis is performed on the compressed data (step S701), MQ decoding (step S702), inverse quantization (step S703), bit shift (step S704), inverse wavelet transform (step S705), and image Is restored (step S706). As described above, in the bit shift method, it is necessary to provide a circuit for generating ROI mask information at the time of encoding and decoding, a bit shift process for wavelet coefficients, and a circuit for calculating a shift amount. In the case of a multi-component image such as a color image, sub-sample processing is often used. If subsamples are used, the ROI mask must be subsampled accordingly.

  As described above, in the above-described prior art, in order to provide an improvement in the image quality of the region of interest while adjusting the code amount so as to become the target code amount, and to realize an effect similar to the ROI function, the bit by the coefficient weighting method is used. There is a problem that the circuit configuration becomes complicated because it is necessary to perform shift and shift amount calculation.

  In order to solve the above problems, for example, Patent Document 5 proposes an image compression apparatus, an image compression method, and a program therefor capable of realizing an image effect similar to ROI with a simple mechanism. In the image compression apparatus or the like, one truncation table is selected, one piece of code block information is read from the memory, the truncation amount is set to 0 if the code block is in the designated area, and the code block is included in the total code amount. If it is outside the designated area, the code amount of the code block when truncation is calculated is added to the total code amount. Do this for all code blocks. Next, one piece of code block information is read. If it is within the specified area, the truncation amount is set to 0 and the code block information is left as it is. If it is outside the specified area, truncation is performed, and the code block information reflecting it is obtained. Write to.

  Further, in Patent Document 6, regions of interest ROI1 to ROI5 are set in an image G1 in order to provide a digital aperture system that can realize a focus effect without using the aperture function and can significantly reduce the data processing amount and processing time. A relatively large amount of code is assigned to the regions of interest ROI1 to ROI5 of the image G1, and the focus is relatively adjusted, and the amount of code is relative to the non-region of interest (non-ROI) 21 of the image G1. The image is compressed by assigning a small amount to the image and relatively blurring the focus, and assigning a relatively small amount of code to the region of interest having a lower priority among the regions of interest ROI1 to ROI5 and making the focus relatively weakly blurred. Then, digital aperture processing is performed on the image G1.

  Further, in Patent Document 7, in order to obtain an image processing apparatus capable of realizing a reduction in system scale and a reduction in processing time in comparing images, in a multi-function machine that is an image processing apparatus that processes image data. The feature amount is extracted for each of the plurality of regions from the first image and the second image that are each divided into a plurality of regions and compressed, and based on the feature amount extracted for each of the plurality of regions from the images. Thus, the first image is compared with the second image. This eliminates the need for processing to divide the image into multiple regions, etc., and makes it possible to perform image comparison with high accuracy with simple processing. As a result, the system scale and processing time are reduced. can do.

JP 2004-304238 A. Japanese Patent Application Laid-Open No. 2004-297712. Japanese Patent Application Laid-Open No. 2004-214828. JP 2007-251476 A. JP 2007-251476 A. JP 2006-295299 A. Japanese Patent Application Laid-Open No. 2004-227406. Ichiro Ueno et al., “Outline of New International Standard System for Still Image Coding (JPEG200)”, Journal of the Institute of Image Information and Television Engineers, Vol. 54, no. 2, pp. 164-171, February 2000.

  However, these conventional methods have the following problems.

  JPEG2000 has a function called ROI (Region of Interest) that can improve the image quality of only a selected portion of an image (see FIG. 1). Usually, in order to realize the ROI function, it is necessary to perform processing corresponding to the encoding (see FIG. 2). As a process, the JPEG 2000 basic method employs a method of weighting the coefficient of the region of interest (specifically, a power of 2 is applied by bit shift). The shift amount of the bit shift is the maximum number of bits of the coefficient that is not the ROI portion, and from this point, it is called a Maxshift system.

  On the other hand, there is also a method for adjusting the amount of code inside and outside the attention area for each code block in the rate control processing part in the process of code formation and realizing an image effect similar to ROI. While this method can achieve an image effect similar to ROI with a simpler mechanism without using a complicated calculation circuit, a code is used instead of code amount calculation for each subband when a rate control table is selected. Since code amount calculation is performed in units of blocks, depending on the number and size of images and regions of interest, there may be disadvantages such as increase in processing time such as code amount calculation and the number of accesses to the external storage device.

  In order to eliminate the above disadvantages, an object of the present invention is to provide an image compression apparatus and method capable of improving the calculation efficiency of the code amount when selecting a rate control table having a ROI attention area, and a program therefor. It is in.

  For distinction, the term “designated region” or “region of interest” is used in place of “ROI” below.

The image compression apparatus according to the first aspect of the present invention relates to the number of coding passes for each subband generated by the MQ encoder, the code amount information for the subbands, the number of coding passes for each code block, and the number thereof. In an image compression apparatus provided with control means for controlling the code amount of a subband with a region of interest, with the code amount information for
The control means determines, based on the usage standard for each subband, which code amount information of the two pieces of code amount information is used, or which code amount information is used, and the code when truncation is performed. It is characterized by calculating the quantity.

  In the image compression apparatus, the control means uses the number and size of specific code blocks as a use standard for each subband.

  Instead, the control means uses the number of accesses to the external storage means as a usage standard for each subband.

  In the image compression apparatus, when the control unit controls the code amount of the subband, when using the two pieces of code amount information, the control unit sets the number of coding passes for each subband and the code amount information corresponding thereto. Based on this, in order to make the image quality in the entire region of interest higher than the image quality outside the region of interest, the number of coding passes for each code block in the region of interest and the code amount information corresponding thereto are added to obtain the code amount of the code block in the region of interest. The code amount of the code block outside the attention area is increased.

  Further, in the image compression apparatus, when the control unit uses two pieces of information when controlling the code amount of the subband, based on the number of coding passes for each subband and the code amount information corresponding thereto, In order to make the image quality in the entire attention area lower than the image quality outside the attention area, the code amount of the code block in the attention area is out of the attention area by subtracting the number of coding passes for each code block in the attention area and the corresponding code amount information. It is characterized by being made smaller than the code amount of the code block.

  Still further, in the image compression apparatus, when the control unit uses the two pieces of code amount information when controlling the code amount of the subband, the control unit sets the number of coding passes per subband and the code amount information corresponding thereto. Based on the above, in order to make the image quality in a part of the attention area higher than the image quality outside the attention area, the code number of the code block in the attention area is added by adding the number of coding passes for each code block in the attention area and the code amount information corresponding thereto. The number of coding passes for each individual code block in the region of interest and the code amount information for it are increased in order to increase the amount more than the code amount of the code block outside the region of interest and lower the image quality in other regions of interest than the image quality outside the region of interest. By subtracting, the code amount of the code block in the attention area is reduced from the code amount of the code block outside the attention area.

In the image compression method according to the second aspect of the invention, the number of coding passes for each subband generated by the MQ encoder, the code amount information for the subbands, the number of coding passes for each code block, and the number thereof are as follows. In an image compression method including a control step for controlling the code amount of a subband having a region of interest, with the code amount information for
The control step determines the code amount when truncating by determining which code amount information of the two pieces of code amount information or both of the code amount information is used based on the use standard for each subband. It is characterized by calculating the quantity.

  In the image compression method, the control step uses the number or size of the specific code block as a use standard for each subband.

  Instead, the control step uses the number of accesses to the external storage means as a usage criterion for each subband.

  In the image compression method, when the code amount of the subband is controlled, the control step uses the number of coding passes for each subband and the code amount information for the subband when using the two pieces of code amount information. Based on this, in order to make the image quality in the entire region of interest higher than the image quality outside the region of interest, the number of coding passes for each code block in the region of interest and the code amount information corresponding thereto are added to obtain the code amount of the code block in the region of interest. The code amount of the code block outside the attention area is increased.

  Furthermore, in the image compression method, when the control step uses two pieces of information when controlling the code amount of the subband, based on the number of coding passes for each subband and the code amount information corresponding thereto, In order to make the image quality in the entire attention area lower than the image quality outside the attention area, the code amount of the code block in the attention area is out of the attention area by subtracting the number of coding passes for each code block in the attention area and the corresponding code amount information. It is characterized by being made smaller than the code amount of the code block.

  Still further, in the image compression method, when the code amount of the subband is controlled, the control step uses the number of coding passes per subband and the code amount information corresponding thereto when using the two pieces of code amount information. Based on the above, in order to make the image quality in a part of the attention area higher than the image quality outside the attention area, the code number of the code block in the attention area is added by adding the number of coding passes for each code block in the attention area and the code amount information corresponding thereto The number of coding passes for each individual code block in the region of interest and the code amount information for it are increased in order to increase the amount more than the code amount of the code block outside the region of interest and lower the image quality in other regions of interest than the image quality outside the region of interest. By subtracting, the code amount of the code block in the attention area is reduced from the code amount of the code block outside the attention area.

  A program for an image compression method according to a third aspect of the invention is a computer-executable program characterized by including a control step of the image compression method.

  According to the image compression apparatus and method of the present invention, the code amount control means can be selected and applied for each subband in the rate control with the attention area, and the code amount when the rate control table with the ROI attention area is selected. The calculation efficiency can be improved.

  In addition, the number of code amount calculation cycles can be reduced in adjusting the code amount inside and outside the attention area, thereby improving the calculation efficiency of the code amount when the rate control table having the ROI attention area is selected.

  Furthermore, it is possible to reduce the number of accesses to the external storage device in adjusting the code amount inside and outside the attention area, thereby improving the calculation efficiency of the code amount when selecting the rate control table with the ROI attention area.

  Also, the code amount can be calculated efficiently when the code amount in the subband attention area is larger than the code amount outside the attention area.

  Furthermore, it is possible to efficiently calculate the code amount when the code amount of the subband attention area is smaller than the code amount outside the attention area.

  Furthermore, the code amount can be calculated efficiently when the code amount in the attention area is increased or decreased by the attention area in the subband.

  Hereinafter, embodiments according to the present invention will be described with reference to the drawings. In addition, in each following embodiment, the same code | symbol is attached | subjected about the same component.

Outline of the embodiment.
At the time of JPEG2000 encoding processing, as disclosed in Patent Documents 1 and 2 described above, code data that has been subjected to arithmetic encoding processing is discarded from the lower order with respect to a preset target code amount, and the target code amount There is a method called “rate control” that adjusts so as to be (see the rate control circuit 32 in FIG. 13). The embodiment of the present invention realizes an effect similar to ROI by adjusting the amount of data to be discarded inside and outside the designated area when performing rate control instead of the bit shift method. In particular, when selecting a rate control table with ROI attention area (step S103 in FIG. 3), code amount calculation efficiency is improved using not only code block unit code amount information but also subband unit code amount information. It is characterized by providing means for making it happen. Specifically, at the time of JPEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the corresponding code amount information, the number of coding passes for each code block, and the corresponding code amount information are input. In the image compression apparatus that controls the code amount of a subband with a region of interest, which code amount information of the two pieces of code amount information is used based on the usage standard for each subband, or both codes It is characterized in that the code amount when truncation is performed after determining whether to use the amount information is calculated.

Embodiment.
FIG. 1 is a diagram showing a flow of processing according to an embodiment of the present invention including rate control processing similar to ROI processing. In the embodiment of the present invention, as shown in FIG. 1, the image is wavelet transformed (step S1), quantized (step S2), MQ (arithmetic) coded (step S3), and rate control having ROI-like functions is performed. Perform (step S4), code formation (step S5), and obtain compressed image data (step S6).

  Next, a rate control procedure according to an embodiment of the present invention will be briefly described.

  FIG. 2 is a block diagram showing an embodiment of reading the code block information. At the time of JPEG2000 encoding, the data after arithmetic encoding processing is the information of each code block (the number of zero bit planes, the number of coding passes, the code block code amount, the code amount for each coding pass, etc.) is an internal memory 64 called a packet pointer. Recorded in the area. The information of one code block forms one packet pointer. The memory address of the packet pointer of each code block is uniquely determined by the component in which the code block indicated by 61 in FIG. 2 is located, the resolution level, the subband information, and the position information of the intra-subband code block indicated by 62. (63). In addition, the code amount is grouped for each coding pass in units of subbands and recorded in the internal memory 64. These pieces of information are read out during rate control.

  If the memory start address of a subband belonging to a certain component and resolution level is SA (SUB), and the vertical and horizontal position of the code block in the subband is (Y, X), the code block information is stored on the memory. The start address SA (CB) of the is as follows.

[Equation 1]
[SA (CB) = SA (SUB) + Y × Xh + X]

  Here, Xh is the number of code blocks in the horizontal direction of the subband. The symbol of the adder in FIG. 2 corresponds to the above formula. In the rate control process described below, the address of the packet pointer is calculated from the position of the code block (63), the packet pointer is read from the internal memory 64, and the code block information (the number of coding passes, the number of bytes, etc.) is extracted. (65).

  FIG. 3 is a diagram showing a flow of general rate control processing. Rate control is roughly divided into two processes.

  The first is a process (A) for selecting a rate control table. This is a process (A) of determining one table to be used from a plurality of prepared rate control tables.

  The second is the truncation process (B). This is a process of truncating the coding pass of each code block using the selected truncation table and changing the code amount information.

  In the truncation table, the truncation amount of the coding pass of the code block in the subband is described for each subband. An example of a truncation table with resolution level 5 is shown below.

[Table 1]
―――――――――――――――――――――――――――
i194 c0: 1, 1 1 1, 1 1 1, 1 1 1, 2 2 2, 6 6 11
i194 c1: 1, 2 2 2, 2 2 3, 4 4 5, 6 7 9, 11 11 16
i194 c2: 1, 1 1 2, 2 2 3, 3 3 4, 5 5 7, 9 9 12
―――――――――――――――――――――――――――

Here, the numbers indicate the following.
C0: Component 0. Specifically, the Y component is indicated.
C1: Component 1. Specifically, the Cb component is shown.
C2: Component 2. Specifically, the Cr component is shown.

The numbers in each column are (C0: 5LL, 5HL-LH-HH, 4HL-LH-HH, 3HL-LH-HH, 2HL-LH-HH, 1HL-LH-HH),
(A) 5LL: truncation amount of subband of level 5LL component in coding pass unit (b) 5HL-LH-HH: truncation amount of subband of level 5 HL component, LH component, and HH component (c) ) 4HL-LH-HH: truncation amount in coding pass unit of level 4 HL component, LH component, and HH component (d) 3HL-LH-HH: level 3 HL component, LH component, HH component sub Truncation amount in band coding pass unit (e) 2HL-LH-HH: Truncation amount in coding pass unit in subband of level 2 HL component, LH component and HH component (f) 1HL-LH-HH: level 1 Truncations of sub-band coding pass of HL component, LH component and HH component Representing ® down the amount, respectively.
Such table data is arranged on the memory in ascending order from the start address.

  First, the process (A) for selecting the rate control table will be described. When rate control is started, one truncation table is selected (step S101), the code amount information for each subband coding pass is read from the memory (step S102), and the truncation is performed according to the selected truncation table. The code amount of the band is calculated (step S103), and the process of step S103 is performed for all the subbands. If the calculated code amount reaches the target code amount, this is used as a final truncation table. If the calculated code amount does not reach the target code amount, the process returns to step S101 to select the next table according to the difference from the target value. Thereby, the truncation amount for achieving the target code amount is determined.

  Next, the flow of truncation (B) will be described. Truncation is started (step S105), the information of one code block is read from the memory (step S106), truncation of the coding block coding path is performed according to the selected truncation table (step S107), and the code reflecting it Block information is obtained and written into the memory as new code block information (step S108). When processing for all code blocks is completed (step S109: Y), the rate control is terminated. Thus, in the rate control process, the number of coding passes to be truncated in subband units is the same.

  Next, an embodiment of the present invention will be described using an example in which the entire code amount is adjusted without performing truncation within a designated area determined by a spatial coordinate position at the time of encoding rate control.

  In the embodiment of the present invention, since the ROI similarity effect is realized by adjusting the truncation amount in units of code blocks, no special processing is required at the time of decoding.

  FIG. 4 is a diagram showing the flow of rate control processing when there is a designated area. Hereinafter, the flow of truncation table selection will be described. First, one truncation table is selected in step S201, and one code block information is read from the memory in step S202. Next, if the code block is a code block in the designated area (step S203: Y), the truncation amount is set to 0, and the code amount of the code block is added to the entire code amount (step S205). If the code block is a code block outside the designated area (step S203: N), the code amount of the code block when truncating the coding pass is performed according to the selected truncation table (step S204), and the entire code Add to the quantity (step S205).

  The above processing (steps S201 to S205) is performed for all code blocks. When the total code amount reaches the target code amount or when the last truncation table is reached (step S206: Y), the final code amount is finalized. A typical truncation table. In other cases (step S207: N), the process returns to step S201 to select the next table according to the difference from the target value.

  Next, the truncation flow will be described. Truncation is started in step S208, and one piece of code block information is read from the memory (step S209). If the read code block is a code block in the designated area (step S210: Y), the truncation amount is set to 0 and the code block information is left as it is (step S211). If the code block is a code block outside the designated area (step S210: N), truncation of the coding pass is performed in accordance with the selected truncation table (step S212), and information on the code block reflecting this is obtained. In the next step S213, the obtained information is written on the memory as new code block information. When the above steps S210 to S213 are performed for all code blocks (step S214: Y), the rate control is terminated. Compared with rate control processing without a designated area, the main difference is that different processing is performed depending on whether or not the code block is in the designated area.

  Next, calculation of code blocks belonging to the designated space area will be described. FIG. 5 is a diagram illustrating an example of code blocks belonging to the designated area. The ROI area is a rectangle, and the spatial information is determined by the upper left vertex A and the lower right vertex B. Assuming that the original image has a resolution level of 0, when the resolution level is increased by 1, the size of the corresponding subband is halved. Therefore, when the resolution level of the designated area of the original image is increased by one, the size in the corresponding subband is halved as shown in FIG. That is, with respect to a code block having a power of two, whether the code block belongs to the designated area can be determined by looking at the upper bits of the vertices A and B of the ROI area.

  FIG. 6 is a diagram illustrating an example of calculation of a code block in a designated area. When the image size is “8192 × 8192” or less, the X-direction coordinates can be covered by a 14-bit register. When the resolution level is 1, the size of the corresponding subband is halved, so from the most significant bit (13th bit) to the bit (1st bit) one bit higher than the least significant bit (0th bit) Represents the X-direction coordinate in the subband. On the other hand, when the code block size is 32 (2 to the 5th power), the lower 5 bits (1 to 5) represent the X coordinate in the code block, and the upper 8 bits (6 to 13) represent the X direction coordinate of the code block. Represent. The Y direction is the same as the X direction, and level 1 and higher are analogized. In this way, the position of the code block and the position in the code block can be calculated.

  FIG. 7 is a diagram showing an example of truncation table selection when there is a designated area. When selecting a truncation table, a value obtained by adding a table address offset (TAO) to a table start address TA (ST) from the memory is used as a table address [TA] [ie, TA = TA (ST) + TAO] (step S301). ). The table address offset (TAO) represents an offset value from the table start address TA (ST) for each table. One table (TBL) is read from the memory based on the table start address TA and recorded in the register (step S302).

  Next, based on the packet pointer address of the code block (CB), the packet pointer of the code block is read from the memory, and the code block information (coding path number CP, code amount BYTE, code amount CP_BYTE (for each coding pass) of the packet pointer is read. N)) is extracted (step S303). If the code block is outside the designated area (step S304: N), the coding pass number (CP) is subtracted by the truncation table value (TBL), and the obtained result is used as a rate-controlled coding pass [CP = CP-TBL] (step S305). Also, based on the code amount information (CP_BYTE (N)) for each coding pass, the code amount from the upper coding pass to the subtracted coding pass is calculated, and the obtained result is used as the code amount after rate control [ TB = TB + CP_BYTE (N)] (step S305).

  On the other hand, if the code block is within the designated area (step S304: Y), the code amount (BYTE) is directly used as the code amount after rate control without performing the subtraction process [TB = TB + BYTE] (step S306). In this way, the calculation is performed for all the components, resolution levels, and subband code blocks, and the resulting code amount (TB) is compared with the target code amount. Select a smaller truncation table.

  Since truncation tables are arranged in ascending order, the table value increases as the address offset (TAO) increases. Initially, the table is selected in the direction of increasing address offset (the width is constant). When the code amount becomes smaller than the set amount, the width is halved, and the table is selected in the direction of decreasing address offset. In this manner, the address offset is updated (TAO update), the search range is gradually narrowed, and the target table is reached. If the target code amount is not reached even when the largest truncation table is selected, the processing ends there, the truncation table at that time is set as the target table, and the code amount at that time is set as the target code amount (step S307 to S308).

  FIG. 8 is a diagram illustrating an example of truncation when there is a designated area. First, the packet pointer of each code block is read, and necessary information such as the number of coding passes and the amount of code (number of bytes) is extracted (step S401). Since the code block in the designated area (step S402: Y) has a truncation amount of 0 and there is no change in the code block information, it is not necessary to perform writing, and the processing proceeds to the next code block (step S408). For the code block outside the designated area (step S402: N), after truncating the coding pass with the truncation table value, the code amount of the resulting number of coding passes is calculated (steps S403 to S406), and the code block information is updated. Therefore, the packet pointer is written into the memory (step S407). When you have finished working on all code blocks, end truncation.

  In this way, the code block in the specified area retains more data (information) than the code block outside the specified area by avoiding data truncation. As a result, it is specified from outside the specified area. The image quality in the region is good, and an effect similar to the ROI function by the bit shift method can be obtained visually, that is, the function similar to the ROI is realized.

  Further, when processing a plurality of frames continuously, rate control (determining a code block using the second code control amount criterion based on a spatial coordinate position designated in advance) is performed for a specific frame. The rate control according to the prior art is performed for other frames. Thereby, it is possible to improve the image quality at a specific position of a specific frame.

  JPEG2000 is a format related to compression / decompression of image data, and hardware resources are required to compress or decompress image data according to this format. This is realized by executing a computer program for image processing by a computer as a hardware resource. In other words, an image processing computer program for image data compression / decompression in JPEG2000 format is installed in a computer storage device, for example, a hard disk drive (HDD), and such an image is created by a basic architecture in which the computer is constructed from a CPU and a memory. It interprets a computer program for processing and executes JPEG 2000 format image compression processing on the input image, or decompresses image data compressed in JPEG 2000 format.

  Computer programs for image processing are marketed via hard disk drives (HDDs), various recording media having optical or magnetic recording information, various portable recording media such as flexible disks, or networks such as the Internet. Can be distributed.

  Next, the code amount calculation method when rate control according to the embodiment of the present invention is examined below.

  There is a method called rate control in which, during the JPEG2000 encoding process, the code data subjected to the arithmetic encoding process is discarded from the lower order with respect to a preset target code amount so as to be adjusted to the target code amount. Further, when performing rate control, the ROI similarity effect can be realized by adjusting the amount of data to be discarded inside and outside the designated area.

  The rate control when there is no ROI region is to calculate the code amount using the code amount information per subband from the MQ encoder, whereas the rate control when there is the ROI region is the code amount for each code block. Therefore, the code amount is calculated using the code amount information for each code block from the MQ encoder. At this time, the code data amount of each code block is read from the external memory and the code data amount to be discarded from the lower order is calculated, so that the calculation time and the external memory are compared with the case where the code amount information for each subband is used. The number of accesses can be increased.

  The embodiment of the present invention efficiently uses the code amount information for each code block as well as the code amount information for each subband to efficiently reduce the amount of code data discarded in the subband when realizing the ROI similarity effect during rate control. It provides a means for calculation.

  From the above description, when there is no ROI designation area in the truncation table selection process (step S103 in FIG. 3 and steps S204, S210 to S212 in FIG. 4), the code amount information for each subband coding pass is used. On the other hand, when there is an ROI designation area, it can be seen that the code amount information for each coding pass of the code block is used. Hereinafter, the code amount calculation and the number of memory accesses will be described in the case where there is no ROI designation area.

  FIG. 14 is a block diagram showing an example (Y component, 1HL) of code amount information for each subband coding pass according to the embodiment of the present invention. FIG. 15 is a block diagram showing an example of code amount information for each coding pass of the code block according to the embodiment of the present invention.

  In FIG. 4 and FIG. 5, a processing example of the Y component level 1 HL subband will be described assuming that the image size is 1024 × 1024, the resolution level is 3, the code block size is 32, and the truncation table value is 5. For simplicity, the number of coding passes of the intra-subband code block is uniformly 30.

  First, when there is no attention area in case 1, the rate control reads out the code amount information for each subband coding pass from the memory once.

  FIG. 16 is a block diagram showing a calculation example (case 1) of the code amount when there is no region of interest according to the embodiment of the present invention. In FIG. 16, since the value of the truncation table is 5, the value obtained by subtracting the code amount corresponding to the lower five coding passes of the code amount information for each subband coding pass from the total code amount after MQ is the sub code after truncation. This is the code amount of the band. The coding band of the lowest subband is 30 and the five coding paths from the lower order are 30, 29, 28, 27, and 26, respectively. At this time, if one cycle is required to calculate the code amount of one coding pass, a total of five cycles are required.

  Next, when there is a region of interest, the rate control reads out the code amount information for each coding pass of each code block in the subband once from the memory, so the total access count is the number of code blocks in the subband × 1 code block Memory access number = 32 × 32 × 1 = 1024 times.

  Here, the following three cases are calculated.

  First, in case 2.1, it is assumed that the number of attention areas is 1, the number of code blocks in the attention area is 1, and the adjustment amount of the number of coding passes in the attention area is a value −3 of the truncation table.

  FIG. 17 is a block diagram illustrating a code amount calculation example (case 2.1) based only on code block information when there is a region of interest according to the embodiment of the present invention. In FIGS. 17 to 22, each block is a code block corresponding to one ROI, the horizontal plane is a code block corresponding to a two-dimensional image, and the uppermost code block is an MSB (most significant). , And the lowest code block corresponds to MLB (lowest).

  In FIG. 17, since the truncation table value is 5, the code amount after truncation of the subband is the remaining coding pass when the number of coding passes of each code block is discarded from the lower 5-3 = 2 passes within the region of interest. Is the total amount of codes for. When the number of coding passes of each code block in the other code blocks is discarded from the lower 5 passes, the total code amount for the remaining coding passes is obtained.

  For example, the lower two coding passes 30 and 29 in the code block in the region of interest are discarded, and the total code amount for each coding pass from 28 to 1 becomes the code amount after truncation of the code block. The lower 30, 30, 28, 27, and 26 coding passes are discarded in the code block outside the region of interest, and the total amount of code for each coding pass from 25 to 1 is the code amount after truncation of the code block. . At this time, assuming that one cycle is required to calculate the code amount of one coding pass of one code block, 30−2 = 28 to calculate the code amount after truncation in each code block in the region of interest. In order to calculate the code amount after truncation of other code blocks, it is necessary to calculate 30−5 = 25 cycles. A total of 1 × 28 + (32 × 32−1) × 25 = 25603 cycles is required to calculate the code amount after truncation of all code blocks in the subband.

  Next, in case 2.2, the number of attention areas is 2, the number of code blocks in attention area A is 200, and the adjustment amount of the number of coding passes in attention area A is a truncation table value −4. It is assumed that the number of code blocks in the attention area B is 300, and the adjustment amount of the number of coding passes in the attention area B is the truncation table value +2.

  FIG. 18 is a block diagram illustrating a code amount calculation example (case 2.2) based only on code block information when there is a region of interest according to the embodiment of the present invention. In FIG. 18, the truncation table value is 5 in the attention area A, and the code amount after truncation of the subband is the code block in the attention area when the number of coding passes is 5-4 = 1 from the lower order is discarded. This is the total code amount for the remaining coding passes.

  In the attention area B, the truncation table value is 5, and the code amount after truncation of the subband is the remaining coding path when the number of coding passes in the code block in the attention area is discarded from the lower 5 + 2 = 7 paths. Is the total amount of codes for. When the number of coding passes in the other code blocks is discarded, the code amount for the remaining coding passes becomes the sum of the coding passes from the lower five. For example, the lower one coding pass 30 in the code block in the attention area A is discarded, and the total code amount for each coding pass from 29 to 1 becomes the code amount after truncation of the code block. The lower seven coding passes 30 to 24 in the code block in the attention area B are discarded, and the total code amount for each coding pass from 23 to 1 is the code amount after truncation of the code block.

  In the code block outside the region of interest, the lower five coding passes 30 to 26 are discarded, and the total code amount for each coding pass from 25 to 1 is the code amount after truncation of the code block. At this time, assuming that one cycle is required to calculate the code amount of one coding pass of one code block, a total of 30 code amounts are calculated to calculate the code amount after truncation for one code block in the region of interest A. -1 = 29 cycles, in order to calculate the code amount after truncation for one code block in the attention area B, a total of 30-7 = 23 cycles, the code amount after truncation of one code block outside the attention area A total of 25 cycles are required to calculate. A total of 200 × 29 + 300 × 23 + (32 × 32−500) × 25 = 25800 cycles is required to calculate the code amount after truncation of all code blocks in the subband.

  Next, in case 2.3, the truncation table value is 26, the number of attention areas is 2, the number of code blocks in attention area A is 200, and the adjustment amount of the number of coding passes in attention area A is the truncation table value −4. . It is assumed that the number of code blocks in the attention area B is 800, and the adjustment amount of the number of coding passes in the attention area B is the truncation table value +2.

  FIG. 19 is a block diagram illustrating a code amount calculation example (case 2.3) based only on code block information when there is a region of interest according to the embodiment of the present invention. In FIG. 19, the value of the truncation table is 26 in the attention area A, and the code amount after truncation of the subband is when the number of coding passes in the code block in the attention area is discarded from the lower 26-4 = 22 paths. This is the total code amount for the remaining coding passes.

  In the attention area B, the truncation table value is 26, and the code amount after truncation of the subband corresponds to the remaining coding paths when the number of coding passes in the code block in the attention area is discarded from the lower 26 + 2 = 28 paths. This is the total code amount. When the number of coding passes in the other code blocks is discarded from the lower 26 coding passes, the total amount of codes for the remaining coding passes is obtained. For example, the lower 22 coding passes 30 to 9 in the code block in the attention area A are discarded, and the total code amount for each coding pass from 8 to 1 becomes the code amount after truncation of the code block.

  The lower 28 coding passes 30 to 3 in the code block in the attention area B are discarded, and the total code amount for each coding pass from 2 to 1 becomes the code amount after truncation of the code block. The lower 26 coding passes 30 to 5 in the code block outside the region of interest are discarded, and the total code amount for each coding pass from 4 to 1 is the code amount after truncation of the code block. . At this time, assuming that one cycle is required to calculate the code amount of one coding pass of one code block, a total of 30 code amounts are calculated to calculate the code amount after truncation for one code block in the region of interest A. -22 = 8 cycles, in order to calculate the code amount after truncation for one code block in the attention area B, a total of 30-28 = 2 cycles, and the code amount after truncation of one code block outside the attention area A total of 30-26 = 4 cycles are required to calculate. A total of 200 × 8 + 800 × 2 + (32 × 32−1000) × 4 = 3296 cycles is required to calculate the code amount after truncation of all code blocks in the subband.

  FIG. 23 is a diagram showing a comparison result of the code amount calculation method, the number of memory accesses, and the code amount calculation cycle number when there is no attention area and when there is no attention area according to the embodiment of the present invention. As is clear from the comparison result of FIG. 23, in order to adjust the image quality inside and outside the attention area when there is the attention area, it is necessary to adjust the number of coding passes and the code amount in units of code blocks, so there is no attention area. Compared to the processing in units of subbands, it can be seen that both the number of memory accesses and the number of code calculation cycles are in an increasing direction, and the throughput of system processing is significantly reduced.

  In the embodiment of the present invention, when there is a region of interest, the code number information for each subband coding pass and the code amount information for each subband coding pass are used in addition to the code amount information for each coding pass of each code block in the subband. The use of two means can be selectively switched for each subband so as to reduce the number of quantity calculation cycles. The mechanism will be described below.

  First, consider the case 2.1 (the number of regions of interest is 1, the number of code blocks in the region of interest is 1, the amount of adjustment of the number of coding passes in the region of interest is the truncation table value-3).

  FIG. 20 is a block diagram illustrating a code amount calculation example (case 2.1) based on subbands and code block information when there is a region of interest according to the embodiment of the present invention.

  When selecting the truncation table, it is necessary to read out the code amount information for each subband coding pass from the memory once. Since the code amount information for each code block coding pass in the subband attention area is read once from the memory, a total of 1 × 1 = 1 memory access is required. Accordingly, a total of 1 + 1 = 2 memory accesses are required when the subband truncation table is selected. On the other hand, since the number of calculation cycles is calculated based on the truncation table value from the subband code amount, the truncation table value is 5, so the number of necessary calculation cycles is 1 × 5 = 5. Times. Next, since the code amount for the number of coding passes when truncation is performed with the truncation value taking into account the adjustment amount (−3) for each code block coding pass in the region of interest, the number of necessary calculation cycles is 1 × (5- 3) = 2 times. The total number of calculation cycles required is 5 + 2 = 7.

  Next, in case 2.2 (the number of attention areas is 2, the number of code blocks in attention area A is 200, and the amount of adjustment of the number of coding passes in attention area A is a truncation table value −4. Consider the case where the number of code blocks is 300 and the adjustment amount of the number of coding passes in the attention area B is the value of the truncation table + 2).

  FIG. 21 is a block diagram illustrating a code amount calculation example (case 2.2) based on subbands and code block information when there is a region of interest according to the embodiment of the present invention.

  When selecting the truncation table, it is necessary to read out the code amount information for each subband coding pass from the memory once. Since the code amount information for each code block coding pass in the subband attention area A is read once from the memory, a total of 1 × 200 = 200 memory accesses are required. Since the code amount information for each code block coding pass in the subband attention area B is read once from the memory, a total of 1 × 300 = 300 memory accesses are required. Therefore, a total of 200 + 300 + 1 = 501 memory accesses are required when selecting the truncation table for the subband. On the other hand, since the number of calculation cycles required first calculates the code amount when truncation is performed using the truncation table value from the code amount of the subband, the number of calculation cycles required is 1 × 5 = 5.

  Next, since the attention area A adds the code amount with respect to the number of coding passes when truncation is performed with the truncation value in consideration of the adjustment amount (−4) for the code block coding pass, the number of code block calculation cycles is (5 -4) = 1. The total number of calculation cycles required for all code blocks in the attention area A is 200 × 1 = 200 times. Since the attention area B subtracts the code amount with respect to the number of coding passes when truncation is performed with a truncation value in consideration of the adjustment amount (+2) for each code block coding pass, the number of code block calculation cycles is (5 + 2) = 7 times. The total number of calculation cycles required for all code blocks in the attention area B is 300 × (5 + 2) = 2100. Therefore, a total of 5 + 200 + 2100 = 2305 calculation cycles are required when selecting the truncation table for the subband.

  Next, in case 2.3 (the truncation table value is 26, the number of attention areas is 2, the number of code blocks in attention area A is 200, and the amount of adjustment of the number of coding passes in attention area A is the truncation table value. -4 Consider the case where the number of code blocks in the attention area B is 800 and the adjustment amount of the number of coding passes in the attention area B is the value of the truncation table +2).

  FIG. 22 is a block diagram illustrating a code amount calculation example (case 2.3) based on subbands and code block information when there is a region of interest according to the embodiment of the present invention.

  When selecting the truncation table, it is necessary to read out the code amount information for each subband coding pass from the memory once. Since the code amount information for each code block coding pass in the subband attention area A is read once from the memory, a total of 1 × 200 = 200 memory accesses are required. Since the code amount information for each code block coding pass in the subband attention area B is read once from the memory, a total of 1 × 800 = 800 memory accesses are required. Therefore, a total of 200 + 800 + 1 = 1001 memory accesses are required when selecting the truncation table for the subband. On the other hand, since the number of calculation cycles required first calculates the code amount when truncation is performed using the truncation table value from the code amount of the subband, the number of necessary calculation cycles is 1 × 26 = 26.

  Next, since the attention area A adds the code amount with respect to the number of coding passes when truncation is performed with the truncation value in consideration of the adjustment amount (−4) with respect to the code block coding pass, the number of code block calculation cycles is (26 -4) = 22 times. The total number of calculation cycles required for all code blocks in the attention area A is 200 × 22 = 4400. The attention area B subtracts the code amount relative to the number of coding passes when truncation is performed with a truncation value that takes into account the adjustment amount (+2) for each code block coding pass, so the number of code block calculation cycles is (26 + 2) = 28 times. The total number of calculation cycles required for all code blocks in the attention area B is 800 × 28 = 22400. Therefore, a total of 26 + 4400 + 22400 = 26826 calculation cycles are required when selecting the truncation table for the subband.

  FIG. 24 is a diagram showing a comparison result of the code amount calculation method, the memory access count, and the code amount calculation cycle number when the code amount calculation method of (subband + code block) is added to the comparison result of FIG.

  As apparent from FIG. 24, in the embodiment of the present invention, when the rate control truncation table is selected with the attention area, the number of memory accesses depends on the total number of code blocks in the subband attention area. When the code amount is calculated by using the code amount information for each code block coding pass as the adjustment amount based on the code amount information for each sub-band coding pass when the number of code blocks in the attention area is small The decrease in the number of times is remarkable. Compared with the code amount calculation using this method and the code amount calculation using only the code amount information for each code block coding pass when the number of subband attention regions is large or the number of code blocks in the subband attention region is large. Thus, the effect of reducing the number of memory accesses is reduced.

  On the other hand, since the number of necessary calculation cycles depends on the value of the truncation table, the number of attention areas, the number of code blocks in the attention area, and the adjustment amount of the code block coding pass in the attention area, the value of the truncation table and the number of attention areas, When the number of code blocks in the region of interest and the code block coding pass adjustment amount (product of both) are small, the code amount information for each code block coding pass is used as the adjustment amount based on the code amount information for each subband coding pass. Thus, when the code amount is calculated, the memory access frequency is remarkably reduced. When the value of the truncation table is large, or when the number of regions of interest and the number of code blocks in the region of interest and the adjustment amount (product of both) are large, only the code amount calculation by this method and the code amount information for each code block coding pass are obtained. Compared with the calculation of the code amount to be used, the effect of reducing the number of memory accesses is reduced.

  Thus, when there is a region of interest and the rate control truncation table is selected, the number of memory accesses and the number of necessary calculation cycles are calculated in advance for each subband, and the code based on the code amount information for each code block coding pass is calculated. If the code amount information for each code block coding pass is adjusted based on the code amount information for each subband coding pass, the code amount is set as an adjustment amount (increase or decrease). Can be calculated.

  In this embodiment, at the time of PEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the corresponding code amount information, the number of coding passes for each code block, and the corresponding code amount information are input. As shown in FIG. 24, when controlling the code amount of a subband with a region of interest, which code amount information of the two pieces of code amount information is used based on the usage standard for each subband. Whether or not both pieces of code amount information are used is determined, and the code amount when truncation is calculated. Here, it is preferable to use the number or size of the specific code block or the number of accesses to the external storage means as the use standard for each subband. Thereby, in the rate control with the attention area, the code amount control means can be selectively applied for each subband, and the calculation efficiency of the code amount can be improved when the rate control table with the ROI attention area is selected.

  In addition, when controlling the code amount of the subband, when using the two pieces of code amount information, attention is paid to the image quality within the entire region of interest based on the number of coding passes for each subband and the code amount information corresponding thereto. Add the number of coding passes for each individual code block in the region of interest and the code amount information to increase the code amount of the code block in the region of interest more than the code amount of the code block outside the region of interest in order to make it higher than the image quality outside the region. Thus, the amount of codes can be calculated more efficiently than in the prior art.

  Furthermore, when controlling the code amount of the subband, when using two pieces of information, the image quality within the entire region of interest is set to be out of the region of interest based on the number of coding passes for each subband and the code amount information corresponding thereto. By subtracting the number of coding passes for each individual code block in the region of interest and the code amount information for it in order to lower the image quality, the code amount of the code block in the region of interest is reduced from the code amount of the code block outside the region of interest, The code amount can be calculated efficiently.

  Furthermore, when controlling the code amount of the subband, when using the two pieces of code amount information, the image quality in a part of the attention area is determined based on the number of coding passes for each subband and the code amount information corresponding thereto. In order to make the image quality outside the region of interest higher than the code amount of the code block outside the region of interest by adding the number of coding passes for each individual code block within the region of interest and the code amount information for the code pass. In order to increase the image quality in other regions of interest to be lower than the image quality outside the region of interest, the code amount of code blocks in the region of interest is subtracted from the number of coding passes for each individual code block in the region of interest and the corresponding code amount information. Is reduced from the code amount of the code block outside the region of interest, so that the code amount can be calculated efficiently.

  As described above, according to the image compression apparatus and method according to the embodiment of the present invention, the code amount control means can be selectively applied for each subband in the rate control with the attention area, and there is the ROI attention area. The code amount calculation efficiency can be improved when the rate control table is selected.

  In addition, the number of code amount calculation cycles can be reduced in adjusting the code amount inside and outside the attention area, thereby improving the calculation efficiency of the code amount when the rate control table having the ROI attention area is selected.

  Furthermore, it is possible to reduce the number of accesses to the external storage device in adjusting the code amount inside and outside the attention area, thereby improving the calculation efficiency of the code amount when selecting the rate control table with the ROI attention area.

  Also, the code amount can be calculated efficiently when the code amount in the subband attention area is larger than the code amount outside the attention area.

  Furthermore, it is possible to efficiently calculate the code amount when the code amount of the subband attention area is smaller than the code amount outside the attention area.

  Furthermore, the code amount can be calculated efficiently when the code amount in the attention area is increased or decreased by the attention area in the subband.

  In the above embodiment, the code amount control method for the image compression method has been described. However, a program that can execute the steps of these methods by a computer is configured, and the program is stored in a CD-ROM, a DVD, or the like. The information may be recorded on a computer-readable recording medium.

  As described above in detail, according to the image compression apparatus and method of the present invention, in the rate control with the attention area, the code amount control means can be selectively applied for each subband, and the rate control with the ROI attention area. The code amount calculation efficiency can be improved when selecting a table.

It is a figure which shows the flow of the process which concerns on embodiment of this invention including the rate control process similar to ROI process. It is a block diagram which shows the Example of reading code block information. It is a figure which shows the flow of a general rate control process. It is a figure which shows the flow of the rate control process when there exists a designated area | region. It is a figure which shows the example of the code block which belongs to the designation | designated area | region. It is a figure which shows the example of calculation of the code block in a designated area. It is a figure which shows the Example of truncation table selection when there exists a designated area | region. It is a figure which shows the Example of truncation when there exists a designated area | region. It is a functional block diagram for demonstrating the basics of a JPEG2000 algorithm. It is a figure which shows the structure of the image compression apparatus disclosed by patent document 2, 3. FIG. It is a figure which shows the flow of the encoding process of JPEG2000 based on a prior art. It is a figure which shows the flow including the ROI process performed at the time of encoding. It is a figure which shows the flow of the ROI process at the time of decoding. It is a block diagram which shows an example (Y component, 1HL) of the code amount information for every coding pass of the subband which concerns on embodiment of this invention. It is a block diagram which shows an example of the code amount information for every coding pass of the code block which concerns on embodiment of this invention. It is a block diagram which shows the calculation example (case 1) when there is no attention area which concerns on embodiment of this invention. It is a block diagram which shows the calculation example (case 2.1) of the code amount only by code block information when there exists an attention area which concerns on embodiment of this invention. It is a block diagram which shows the calculation example (case 2.2) of the code amount only by code block information when there exists an attention area which concerns on embodiment of this invention. It is a block diagram which shows the calculation example (case 2.3) of the code amount only by code block information when there exists an attention area which concerns on embodiment of this invention. It is a block diagram which shows the calculation example (case 2.1) of the code amount by a subband and code block information when there exists an attention area which concerns on embodiment of this invention. It is a block diagram which shows the calculation example (case 2.2) of the code amount by a subband and code block information when there exists an attention area which concerns on embodiment of this invention. It is a block diagram which shows the calculation example (case 2.3) of the code amount by a subband and code block information when there exists an attention area which concerns on embodiment of this invention. It is a figure which shows the comparison result of the code amount calculation method, the memory access frequency, and the code amount calculation cycle number with and without the attention area according to the embodiment of the present invention. FIG. 24 is a diagram illustrating a comparison result of a code amount calculation method, a memory access count, and a code amount calculation cycle number when a code amount calculation method of (subband + code block) is added to the comparison result of FIG.

Explanation of symbols

10: Wavelet transform unit,
20: arithmetic coding unit,
30 ... Packet header generation processing unit,
40 ... Memory controller,
50 ... DRAM,
13, 21, 25, 31, 33, 35, 37 ... DMA,
11 ... color conversion circuit,
12 ... Wavelet transform circuit,
23: Bit plane dividing circuit,
24. Arithmetic coding circuit,
22: Quantization circuit,
26: Average value calculation circuit,
27. Masking coefficient calculation circuit,
28: Code amount calculation circuit,
29: Data processing circuit,
32. Rate control circuit,
34 ... Packet information generation circuit,
36 ... Packet header generation circuit,
38 ... code forming circuit,
100: Image compression device,
111 ... color space conversion / inverse conversion unit,
112 ... 2D wavelet transform / inverse transform unit,
113 ... Quantization / inverse quantization unit,
114 ... entropy encoding / decoding unit,
115: Tab processing unit.

Claims (7)

At the time of JPEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the code amount information for the subbands, the number of coding passes for each code block and the code amount information for the code region information are input. In an image compression apparatus provided with rate control means for controlling the code amount of a certain subband,
The rate control means, when controlling the code amount of the subband, when using the two code amount information, based on the number of coding passes for each subband and the code amount information corresponding thereto, In order to make the image quality of the code block outside the region of interest code by adding the number of coding passes for each code block within the region of interest and the code amount information for that code An image compression apparatus characterized in that the image compression apparatus increases the amount . At the time of JPEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the code amount information for the subbands, the number of coding passes for each code block and the code amount information for the code region information are input. In an image compression apparatus provided with rate control means for controlling the code amount of a certain subband,
The rate control means, when controlling the code amount of the subband, when using the two pieces of information, based on the number of coding passes for each subband and the code amount information corresponding thereto, the image quality within the entire region of interest In order to lower the image quality outside the region of interest, the number of coding passes for each individual code block in the region of interest and the code amount information for the code pass are subtracted to obtain the code amount of the code block in the region of interest from the code amount of the code block outside the region of interest An image compression apparatus characterized in that the image compression apparatus is reduced . At the time of JPEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the code amount information for the subbands, the number of coding passes for each code block and the code amount information for the code region information are input. In an image compression apparatus provided with rate control means for controlling the code amount of a certain subband,
The rate control means, when controlling the code amount of the subband, when using the two code amount information, based on the number of coding passes per subband and the corresponding code amount information, In order to make the image quality within the region of interest higher than the image quality outside the region of interest, the number of coding passes for each code block within the region of interest and the code amount information corresponding thereto are added, and the code amount of the code block within the region of interest is In order to increase the code amount and make the image quality in other regions of interest lower than the image quality outside the region of interest, the number of coding passes for each code block in the region of interest and the code amount information corresponding thereto are subtracted and the code block within the region of interest The image compression apparatus is characterized in that the code amount is reduced from the code amount of the code block outside the region of interest . At the time of JPEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the code amount information for the subbands, the number of coding passes for each code block and the code amount information for the code region information are input. In an image compression method including a rate control step for controlling a code amount of a certain subband,
In the rate control step, when controlling the code amount of the subband, when using the two pieces of code amount information, based on the number of coding passes for each subband and the corresponding code amount information, In order to make the image quality of the code block outside the region of interest code by adding the number of coding passes for each code block within the region of interest and the code amount information for that code A method of compressing an image, wherein the image compression method increases the amount . At the time of JPEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the code amount information for the subbands, the number of coding passes for each code block and the code amount information for the code region information are input. In an image compression method including a rate control step for controlling a code amount of a certain subband,
In the rate control step, when controlling the code amount of the subband, when using the two pieces of information, the image quality within the entire region of interest is determined based on the number of coding passes for each subband and the code amount information corresponding thereto. In order to lower the image quality outside the region of interest, the number of coding passes for each individual code block in the region of interest and the code amount information for the code pass are subtracted to obtain the code amount of the code block in the region of interest from the code amount of the code block outside the region of interest. An image compression method characterized by reducing. At the time of JPEG2000 encoding, the number of coding passes for each subband generated by the MQ encoder and the code amount information for the subbands, the number of coding passes for each code block and the code amount information for the code region information are input. In an image compression method including a rate control step for controlling a code amount of a certain subband,
In the rate control step, when controlling the code amount of the subband, when using the two pieces of code amount information, based on the number of coding passes for each subband and the corresponding code amount information, a part of the attention area In order to make the image quality within the region of interest higher than the image quality outside the region of interest, the number of coding passes for each code block within the region of interest and the code amount information corresponding thereto are added, and the code amount of the code block within the region of interest is In order to increase the code amount and make the image quality in other regions of interest lower than the image quality outside the region of interest, the number of coding passes for each code block in the region of interest and the code amount information corresponding thereto are subtracted and the code block within the region of interest The image compression method is characterized in that the code amount is reduced from the code amount of the code block outside the region of interest . A computer-executable program comprising the rate control step of the image compression method according to any one of claims 4 to 6 .

Images