JP2010213059A - Image encoding apparatus and image encoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image encoding apparatus capable of encoding a high-resolution image under a limitation in memory capacity. <P>SOLUTION: The image encoding apparatus performing wavelet transform processing using a coefficient line buffer, comprises: an image input section 201 for inputting an image to be encoded; a code block size determining section 202 for determining the longitudinal size of a code block being the minimum unit of encoding, on the basis of the lateral size of the image input by the image input section 201; and an image encoding section 203 for encoding the image using the size of the code block determined by the code block size determining section 202. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an image encoding apparatus and an image encoding method for encoding an image using the JPEG2000 method, and in particular, when encoding an image, a line-based wavelet transform is used and a memory without reducing compression efficiency. The present invention relates to an image encoding apparatus and an image encoding method for efficiently compressing data under capacity limitations.

  As a still image compression method, the basic method (part 1) of JPEG 2000, which has higher image quality and more functions than the widely used JPEG (Joint Photographic Experts Group), is ISO / IEC 15444-1, ITU- TT 800 is standardized. JPEG2000 is characterized in that the image quality is improved at a high compression rate, the code amount control is easy, the code data can be edited, and the functionality is higher than that of the conventional JPEG.

  The JPEG2000 encoding method employs wavelet transform. The wavelet transform is a frequency transform process applied to the entire image, not a block unit process such as the discrete cosine transform (DCT) used in the conventional JPEG. Patent Document 1 discloses an effective encoding and decoding method using a wavelet transform method. When adopting a method based on wavelet transform, each subband region corresponding to a rectangular region in the original image is set as one encoding unit, and the capacity of the line buffer is determined at the decomposition level.

  Since the wavelet transform is a frequency transform process applied to the entire image, if it is simply implemented, a buffer for holding pixel values or transform coefficients corresponding to the image size is required throughout the process.

  As a solution to this, there is a line-based wavelet processing method as described in Non-Patent Document 2. The line-based wavelet processing is a technique for performing conversion processing by sequentially storing pixel values or conversion coefficients in a minimum necessary buffer (line memory) corresponding to the number of taps of wavelet conversion. After the line-based wavelet transform, these coefficients are not held all at the same time, but are sequentially entropy-coded to make a buffer space for the coefficients. Therefore, this creates the effect of greatly reducing the amount of memory consumed.

  A method for performing wavelet transform on a line basis is also disclosed in Patent Document 2. According to Patent Document 2, it is described that a buffer necessary for wavelet transform is not a buffer corresponding to the entire image but a line buffer having the number of lines that does not depend on the vertical size of the image.

  However, even when a line-based wavelet is used, the required buffer amount, that is, the amount of memory consumed varies depending on the coefficient buffer holding method, the wavelet transform decomposition level, the unit for executing encoding, and the like. Patent Document 2 discloses a method for performing wavelet transform on a line basis, but there is no description regarding control of the code block size, and there is also description regarding a method of compression under the limitation of available buffers. Not.

  In recent years, the resolution of images processed by digital cameras, copiers, etc. has become very high, and even when using line-based wavelets, the memory consumed is much smaller than that of wavelets that process the entire image. The problem remains that the amount of memory consumed increases as the processing in the process increases.

  In addition, in such devices as digital cameras, digital scanners, multifunction devices, etc., not only the installed memory is small, but there are many cases where the memory that can be used is limited, and for any image that can be input, Control is required to operate normally within the available memory capacity.

  The present invention has been made in view of the above-described problems. When an image is encoded by the JPEG2000 system using a line-based wavelet, the memory capacity is limited without reducing the compression efficiency of the image. One of the purposes is to realize efficient image coding.

  In order to solve the above-described problems and achieve the object, an image encoding apparatus according to the present invention is an image encoding apparatus that performs wavelet transform processing using a coefficient line buffer. An image input means for inputting, a code block size determining means for determining a vertical size of a code block which is a minimum unit of encoding based on a horizontal size of an image input by the image input means, and a code block size Image encoding means for encoding an image using the code block size determined by the determining means.

  According to the present invention, since an image is encoded using line-based wavelet transform, when performing image processing, the image can be processed within a usable memory capacity, and memory saving can be realized. In addition, since the code block is appropriately set according to the size of the input image, it is possible to realize efficient image coding under the limitation of the memory capacity without reducing the compression efficiency of the image. There is an effect.

FIG. 1 is a block diagram for explaining basic encoding and decoding algorithms of JPEG2000. FIG. 2 is a diagram illustrating subband coefficients after one-stage subband division (forward conversion) and deinterleaving (inverse conversion). FIG. 3 is a diagram showing the relationship among images, tiles, subbands, precincts, and code blocks in image division. FIG. 4 is a diagram illustrating the coding order of code blocks and bit planes. FIG. 5 is a diagram illustrating an example of a layer and a packet included in the layer when the decomposition level is 2 and the precinct size is equal to the subband size. FIG. 6 is a diagram showing the progression order of packet attribute hierarchies in JPEG2000. FIG. 7 is a block diagram showing a hardware configuration of the PC according to the first embodiment of the present invention. FIG. 8 is a block diagram illustrating an encoding process of the image encoding device according to the first embodiment. FIG. 9 is a diagram illustrating a line-based wavelet window according to the first embodiment. FIG. 10 is a conceptual diagram showing line-based wavelet transform according to the first embodiment. FIG. 11 is a flowchart showing entropy encoding processing by line-based wavelet transform according to the first embodiment. FIG. 12 is a flowchart showing subband decomposition processing to the next stage for the LL subband. FIG. 13 is a functional block diagram relating to wavelet transform and encoding. FIG. 14 is a diagram showing an example of securing a coefficient buffer in line-based encoding according to the present invention. FIG. 15 is a table showing restrictions on the code block size of JPEG2000. FIG. 16 is a diagram illustrating an example of two images having different sizes and their code block sizes. FIG. 17 is a diagram illustrating an example of two images having different sizes and their code block sizes. FIG. 18 is a block diagram illustrating an encoding process of the image encoding device according to the fourth embodiment of the present invention. FIG. 19 is a diagram showing a precinct division unit related to precinct division.

  DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Exemplary embodiments of an image encoding device and an image encoding method according to the present invention are explained in detail below with reference to the accompanying drawings. The present invention is not limited to the embodiments. In addition, constituent elements in the following embodiments include those that can be easily assumed by those skilled in the art or those that are substantially the same.

  An overview of JPEG2000 encoding processing will be described with reference to FIGS. The details of the JPEG2000 conversion formula and the like can be obtained by referring to Non-Patent Document 1. JPEG2000 is one of image compression methods, and is a specification developed from JPEG, which has been widely used in the past, and has been standardized and standardized. As described above, the wavelet transform method is employed, and compression is performed by quantizing and encoding the vertical and horizontal frequency components divided into frequency bands. Specifically, the action of the wavelet is to divide a one-dimensional wave into a low-frequency component and a high-frequency component, and the original signal is restored from the combination of the low-frequency component and the high-frequency component in inverse transformation. When two-dimensional data such as an image is subjected to wavelet transform, the image is regarded as a set of row × column data, and therefore, the transformation is performed for each row or each column.

  FIG. 1 is a block diagram for explaining a basic encoding and decoding algorithm of JPEG2000 according to the present invention.

  First, the image is divided into rectangular regions (tiles) that do not overlap (number of divisions ≧ 1) and processed in units of tiles (the entire image can be processed as one tile). Next, when the tile encodes a color image composed of three RGB components, the divided tiles are subjected to color conversion into luminance and color difference components. The color conversion unit 1 can perform processing for increasing the compression efficiency of a color image by converting the RGB image into a luminance-chrominance YCrCb image. Color conversion is performed in order to efficiently compress a color image. Two methods of reversible and irreversible are defined for color conversion.

  The color-converted component is subjected to wavelet transform in the two-dimensional wavelet transform unit 2 and divided into four subbands described later. In the quantization unit 3, quantization is performed, and encoded data is generated via the entropy encoding unit 4 and the tag processing unit 5. This processing is performed by forward conversion in encoding (compression), and in reverse conversion is performed by decoding (decompression).

  In the JPEG2000 wavelet transform, an integer type 5 × 3 filter and a real number type 9 × 7 filter are employed. The feature of the reversible 5 × 3 filter is that the coefficients used for conversion are integers, and the converted coefficients are converted into integers by rounding.

Here, the conversion formula of the reversible 5 × 3 filter is shown in Formula (1).
Forward conversion
Y (2n + 1) = X (2n + 1) -floor ((X (2n) + X (2n + 2)) / 2) [step 1]
Y (2n) = X (2n) + floor ((Y (2n-1) + Y (2n + 1) +2) / 2) [step 2]
Inverse transformation (1)
X (2n) = Y (2n) -floor ((Y (2n-1) + Y (2n + 1) +2) / 4) [step 1]
X (2n + 1) = Y (2n + 1) + floor ((X (2n) + X (2n + 2)) / 2) [step 2]

In contrast to the reversible 5 × 3 filter, the feature of the irreversible 9 × 7 filter is that the coefficients used for the conversion are real numbers and that the converted coefficients are not rounded. is there. Hereinafter, the conversion formula of the irreversible 9 × 7 filter is shown in Formula (2).
Forward conversion
Y (2n + 1) = X (2n + 1) + α * (X (2n) + X (2n + 2)) [step 1]
Y (2n) = X (2n) + β * (Y (2n-1) + Y (2n + 1)) [step 2]
Y (2n + 1) = Y (2n + 1) + γ * (Y (2n) + Y (2n + 2)) [step 3]
Y (2n) = Y (2n) + δ * (Y (2n-1) + Y (2n + 1)) [step 4]
Y (2n + 1) = Κ * Y (2n + 1) [step 5]
Y (2n) = (1 / Κ) * Y (2n) [step 6]
Inverse transformation (2)
X (2n) = Κ * Y (2n) [step 1]
X (2n + 1) = (1 / Κ) * Y (2n + 1) [step 2]
X (2n) = X (2n) -δ * (X (2n-1) + X (2n + 1)) [step 3]
X (2n + 1) = X (2n + 1) -γ * (X (2n) + X (2n + 2)) [step 4]
X (2n) = X (2n) -β * (X (2n-1) + X (2n + 1)) [step 5]
X (2n + 1) = X (2n + 1) -α * (X (2n) + X (2n + 2)) [step 6]
However, α = -1.586134342059924
β = -0.052980118572961
γ = 0.882911075530934
δ = 0.443506852043971
Κ = 1.230174104914001

  FIG. 2 is a diagram illustrating subband coefficients after one-stage subband division (forward conversion) and deinterleaving (inverse conversion). By the wavelet transform, the original image is divided into four subbands HH, HL, LH, and LL. As described above, in the two-dimensional wavelet transform, the vertical direction conversion and the horizontal direction conversion are sequentially performed. Therefore, each of the two-dimensional wavelet transforms is divided into half sizes, resulting in the formation of four quarter-size subbands. Is done. When the wavelet transform is recursively performed on the LL subband from among the divided four subbands, and the recursive wavelet transform (decomposition) is repeated, finally one subband LL and a plurality of subbands Subbands HL, LH, and HH are generated.

  The coefficients after conversion in the forward conversion are decimated to ½, and the number of coefficients is ½ in both the horizontal direction and the vertical direction with respect to the pre-conversion, but there are four subs of HH, HL, LH, and LL. Since it is decomposed into bands, the number of coefficients does not change before and after conversion. By repeating this decomposition recursively for the LL component, JPEG2000 frequency decomposition is realized.

  FIG. 3 is a diagram illustrating the relationship among images, tiles, subbands, precincts, and code blocks in image division. The precinct is a subband divided into rectangles, and a set of three regions at the same spatial position of subbands HL, LH, and HH having the same decomposition level is treated as one precinct. However, in the LL subband, one area is treated as one precinct. The size of the precinct can be the same as that of the subband, and the rectangular area into which the precinct is divided is a code block. Here, the code block is a unit of encoding, and entropy encoding is performed independently for each code block. The physical size order is image ≧ tile> subband ≧ precinct ≧ code block.

  FIG. 4 is a diagram illustrating an encoding order for a code block and a bit plane. By performing subband division, coefficient entropy coding (bit-plane coding) is performed for each code block and in bit-plane order. A part of the code of the bit plane is extracted from all the code blocks included in the precinct and collected, and the one with the header is called a packet. The extracted code may be, for example, a collection of the code of bit planes up to the third one from the MSB of all code blocks. In addition, since a part of the code may be in an “empty” state, the contents of the packet may be “empty” from a code point of view. The header of the packet includes information regarding the code included in the packet, and each packet can be handled independently. A packet is a unit of code.

  When packets of all precincts (ie, all code blocks and all subbands) are collected, a part of the code of the entire image area (for example, the MSB to the third bit plane of the wavelet coefficients of the entire image area) Is called a layer. Since the layer is a part of the code of the bit plane of the entire image, the image quality of the image becomes better as the number of decoded layers increases. That is, the layer is a unit of image quality formed in the bit depth direction. By collecting all the layers, the bit plane of the entire image is encoded.

  FIG. 5 is a diagram illustrating an example of a layer and a packet included in the layer when the decomposition level is 2 and the precinct size is equal to the subband size. Since the packet is based on the precinct, when the precinct is the same as the subband, the packet spans the subbands HL to HH. Further, in order to clarify the explanation, some packets are surrounded by thick lines in the figure.

  FIG. 6 is a diagram showing the progression order of packet attribute hierarchies in JPEG2000. The operation of arranging packets according to the generated packet and layer division is called code formation. That is, a packet has four attributes, which component belongs to, which resolution level belongs to which precinct (location), and which layer belongs to. The packet arrangement means which attributes are arranged in a hierarchical order. The arrangement order is called a progression order, and referring to FIG. 6, five arrangement orders are shown.

Next, an example of how the encoder arranges packets in the order of progression and how the decoder interprets the attributes of the packets in the order of progression is shown. If the component is C, the resolution level is R, the precinct is P, and the layer is L, in the case of LRCP progression, the following for loop nested in the order of L, R, C, P
for (layer) {
for (resolution level) {
for (component) {
for (Precinct) {
Encoding: Place packet
When decoding: Interpret packet attributes}
}
}
}
In the hierarchical order, packets are arranged at the time of encoding, and packets are interpreted at the time of decoding. Each packet has a packet header. The information included in the packet header includes whether the packet is empty, which code block is included in the packet, the number of zero bit planes in each code block included in the packet, and the code block code included in the packet. The number of coding passes (number of bit planes), the code length of each code block included in the packet, and the like are included.

  However, there is no description of the layer number, resolution number, etc. in the packet header, and in order to determine which layer of which resolution the packet is at the time of decoding, from the progression order described in the COD marker in the main header. The for loop may be formed, a packet break may be determined from the sum of the code lengths of the code blocks included in the packet, and it may be determined at which position in the for loop each packet is handled. That is, as long as the code length in the packet header can be read, the next packet can be detected without decoding the entropy code itself. Therefore, it means that any packet can be accessed.

(First embodiment)
In the first embodiment of the present invention, an image encoding apparatus that performs an image encoding process using a wavelet transform method using a coefficient line buffer in a JPEG2000 encoding method that uses a wavelet transform method will be described. .

  FIG. 7 is a block diagram illustrating a hardware configuration of the personal computer (PC) according to the first embodiment. A PC 100 according to this embodiment is a general PC in which a CPU (Central Processing Unit) 101, a RAM (Random Access Memory) 102, an HDD (Hard Disk Drive) 103, and the like are interconnected by a system bus 104. In this embodiment, a general PC provided with an image encoding device has been described. However, the present invention is not limited to a PC, and image codes such as MFP (Multi Function Peripheral), digital copiers, printers, facsimiles, and copies with scanner functions are used. The image encoding apparatus according to the present embodiment can be applied as long as it can perform the encoding process.

  The CPU 101 is a microprocessor for controlling the entire PC 100. Since the RAM 102 has a property of storing various data in a rewritable manner, the RAM 102 functions as a work area for the CPU 101 and functions as a buffer. The HDD 103 is an information recording medium that stores images and various programs executed by the CPU 101.

  As the information recording medium, not only the HDD 103 but also various optical disks such as CD (Compact Disk) -ROM and DVD (Digital Versatile Disk), various magnetic disks such as various magneto-optical disks and flexible disks, semiconductor memory, etc. Various types of media can be used. Alternatively, the program may be downloaded from a network such as the Internet via an external I / F device (not shown) and installed in the HDD 103. In such a case, the storage device that stores the program on the server on the transmission side is also an information recording medium in the present invention. The program may operate on a predetermined OS (Operating System). In that case, the OS may take over some of the various processes described below. Alternatively, it may be included as part of a group of program files that constitute predetermined application software, OS, or the like.

  The CPU 101 that controls the operation of the entire system executes various processes based on programs stored in the HDD 103 used as the main storage device of the system.

  Next, among the functions that the program stored in the HDD 103 of the PC 100 causes the CPU 101 to execute, the characteristic functions provided in the PC 100 of this embodiment will be described.

  First, with reference to FIG. 7, an outline of characteristic functions provided in the PC 100 will be described. (1) From the HDD 103, an original image to be processed is read into the RAM 102 by a command from the CPU 101. (2) The CPU 101 reads the original image on the RAM 102, determines the code block size of the present embodiment, and performs encoding using wavelet transform, as will be described later. (3) The CPU 101 writes the encoded data in another area on the RAM 102. (4) The encoded data is stored in the HDD 103 according to a command from the CPU 101.

  FIG. 8 is a block diagram illustrating an encoding process of the image encoding device according to the present embodiment. According to FIG. 8, the image input unit 201 inputs an image read into the RAM 102 to be encoded as an input image. The code block size determining unit 202 takes in the size of the image (the size in the horizontal direction of the image) input by the image input unit 201 and, based on the size of the image, the size in the vertical direction of the code block that is the minimum unit of encoding To decide. The image encoding unit 203 inputs an image from the image input unit 201, performs wavelet transform processing on the input image using the code block size determined by the code block size determination unit 202, and encodes the input image. . Here, the encoding process performed in the image encoding unit 203 is a process for performing all encoding steps of quantization, coefficient bit modeling, and entropy encoding. The encoded image is temporarily written in the RAM 102 and written in a predetermined area on the HDD 103.

  Next, line-based encoding according to this embodiment will be described in detail. FIG. 9 is a view showing a line-based wavelet window according to the present embodiment. As described above, the line-based wavelet is different from the method in which the coefficient buffer for the entire image is prepared and wavelet transform is performed. As shown in FIG. 9, the image is cut into strips in the horizontal direction and L in the vertical direction. This is a method of performing wavelet transform in order on a window having a length. Here, it is not necessary for the coefficient buffer to hold the entire image, and it is sufficient to hold coefficient buffers for several lines at the same time.

  FIG. 10 is a diagram illustrating the line-based wavelet transform according to the present embodiment. 10 will be described in detail. First, L-line image data (input pixels in FIG. 10) is read into the line buffer. Here, when color conversion is performed on the pixel value of the input image, the pixel value of each plane after color conversion is developed in a line buffer of L lines. Next, when vertical wavelet transform is performed on the image data, the image data is decomposed into a vertical high-pass component and a vertical low-pass component by a high-pass filter and a low-pass filter. At this time, the value of L needs to be larger than the filter size used for the wavelet transform.

  Next, horizontal wavelet transform is performed on each of the vertical high-pass component and the vertical low-pass component. As a result, the coefficient components of the subbands HH, HL, LH, and LL are generated, which are the same as the subband coefficient components generated when the wavelet transform is simultaneously performed on the entire image. FIG. 10 shows an example in which a 9 × 7 filter is used for wavelet transform.

  If the length of the filter in FIG. 10 is F, the pixel value of the F line (or the pixel value after color conversion) is read, and one-line subbands HH, HL, LH, and LL components are output. For example, when the (F + K) line is read at a time, each subband signal of the (1 + K) line is output at a time. Among the generated subband signals, the HH, HL, and LH subband signals are not further decomposed, and therefore the HH, HL, and LH subband signals proceed to the encoding step (see FIG. 11). On the other hand, among the generated subbands, the LL subband is further subjected to the next wavelet transform to be decomposed into subband signals of lower decomposition levels.

  FIG. 11 is a flowchart showing entropy encoding processing by line-based wavelet transform according to the present embodiment. Here, the line-based encoding processing of HH, HL, and LH subband signals will be briefly described with reference to a flowchart.

  First, L lines following the image cut out in a strip shape from the image are read into the coefficient buffer (step S1101). Wavelet transform is performed on the read coefficient buffer data (step S1102). Here, since the unit for performing entropy coding is for each code block, it is necessary to generate coefficients for the code block. The coefficients subjected to the wavelet transform are stored in a memory until a coefficient for each code block is generated. The coefficients are output on the image line, and if sufficient transform coefficients are generated for the vertical size of the code block (step S1103, Yes), the subband coefficients are quantized (step S1104). The quantized subband coefficients are subjected to bit modeling (step S1105) and subjected to entropy coding (step S1106). On the other hand, if a sufficient transform coefficient is not generated for the code block size (step S1103, No), the process returns to step S1101, and the subsequent coefficient buffer reading process is performed. The above is the process for one-stage wavelet transform.

  FIG. 12 is a flowchart showing the subband decomposition process of the next stage for the LL subband. Hereinafter, the processing of the LL subband will be briefly described. As in the flowchart of FIG. 11, first, L rows cut out in a strip shape from an image are read into the coefficient buffer (step S1201), and wavelet transform is performed on the data in the coefficient buffer (step S1202). Next, as described above, the LL subband signal is further subjected to the next wavelet transform. Therefore, if LL subband coefficients sufficient for re-decomposition are generated (step S1203, Yes), the LL subband signal is generated. A coefficient wavelet transform is performed to decompose the coefficient into lower decomposition levels (step S1204). If an LL subband coefficient sufficient for re-decomposition is not generated (step S1203, No), the process returns to step S1201, and the subsequent coefficient buffer reading process is performed. Here, the LL subband component that has reached the maximum resolution level is encoded according to the flow of the flowchart of FIG. 11 in the same manner as other HH, HL, and LH subbands.

  FIG. 13 is a functional block diagram relating to wavelet transform and encoding. As shown in FIG. 13, the coefficients subjected to wavelet transform and subband decomposition are stored in HH, HL, LH, and LL coefficient buffers. Among the coefficients stored in the coefficient buffer, the HH, HL, and LH coefficients are input to the encoding unit, and a code is generated. For the coefficients of the LL subband, the next wavelet transform is performed.

  Next, an example of securing a coefficient buffer associated with line-based encoding according to the present embodiment will be described. FIG. 14 is a diagram illustrating an example of securing a coefficient buffer associated with line-based encoding in the present embodiment. For example, when the decomposition level is 3, the subband coefficients of a plurality of subbands 1HH, 1HL, 1LH, 1LL, 2HH, 2HL, 2LH, 2LL, 3HH, 3HL, 3LH, and 3LL are obtained in the multistage wavelet transform process. Generated.

  When using wavelet transform that is applied to the entire image instead of line-based wavelet transform, a plurality of subband coefficients are simultaneously developed in the memory. However, according to the line-based wavelet transform, since processing is performed for each code block, it is possible to develop a line corresponding to the vertical size of the code block of each subband at the same time.

  For example, if the code block size is 32 * 32, 32 lines of coefficient buffers are simultaneously held for all subbands. The HH, HL, and LH subbands in which the coefficients for 32 lines are accumulated are subjected to entropy coding at any time, and when the codes are generated, there is no need to hold them in the memory, and the portions are overwritten. Here, it is desirable to use a 32-line circular buffer preferentially overwritten from the old line.

  Further, since the subband coefficients of the 1LL and 2LL subbands are recursively wavelet transformed, 32 lines are not necessarily required. However, since wavelet transformation is further performed, it is necessary to provide a line buffer having a filter size necessary for wavelet transformation. Again, it is desirable to use a circular buffer depending on the window size at the time of wavelet transform.

  When the coefficient buffer described above is used, the amount of memory required in the process of line-based wavelet transform is the size in the horizontal direction (horizontal size) relative to the input direction of the input image and the size in the vertical direction of the code block (vertical size). Determined by.

よって、上記式（３）を用いて、使用可能なメモリ使用量と画像横サイズから、コードブロックの縦方向の最大サイズを算出することが望ましい。 Here, a method for calculating the vertical size of the code block will be described. The horizontal size of the input image is W, the vertical size of the code block is C _v , the accuracy of the input pixel (or pixel after color conversion) is P bytes, the resolution level is n, and the accuracy of the subband coefficients is When Bytes _Coeff bytes are used, the memory usage calculated from the sum of the window line buffer that reads pixel values (or LL components) line by line when performing wavelet transform and the coefficient buffer that holds all subband coefficients at the same time is And can be calculated as shown in Equation (3).

Therefore, it is desirable to calculate the maximum size of the code block in the vertical direction from the usable memory usage and the horizontal size of the image using the above equation (3).

  Here, there is a certain restriction on the code block size. FIG. 15 is a table showing restrictions on the code block size of JPEG2000. As shown in FIG. 15, the code block size does not exceed the calculated maximum code block size and is a power of 2, and the sum of exponents of powers of 2 in the vertical size and the horizontal size is 12 or less. It is desirable to select.

  According to the image coding apparatus according to the present embodiment, since the vertical size of the code block, which is the minimum unit of coding, is determined by the horizontal size of the input image, the image is processed within the usable memory capacity. Thus, memory saving can be realized even for high-resolution images. In addition, since the horizontal size of the input image is used to calculate the vertical size of the code block from the maximum capacity of the memory area that can be used for the wavelet transform process, without reducing the encoding efficiency of the image, Efficient encoding can be realized under the limitation of memory capacity.

(Second Embodiment)
In the second embodiment of the present invention, when determining the vertical size of the code block, if the horizontal size of the second input image is larger than the horizontal size of the first input image, An image coding apparatus that reduces the size of the code block used for the second image in the vertical direction from the size of the code block used for the image will be described.

  Also in this embodiment, a PC having the general hardware configuration described in the first embodiment is used (see FIG. 7). Moreover, it similarly has the function of the encoding process of the image encoding device shown in the first embodiment (see FIG. 8).

  A line-based wavelet transform that uses a coefficient line buffer based on a line obtained by cutting an image into a strip in the horizontal direction. When processing images with very large horizontal sizes, a large amount of work can be done as a wavelet transform coefficient buffer. Memory is required. Therefore, it is desirable to determine the code block size according to the horizontal size of the input image.

  FIG. 16 is a diagram illustrating an example of two images having different sizes and their code block sizes. As shown in FIG. 16, for example, when encoding an image 1 in which the horizontal size of the input image is 2048 pixels, the code block is encoded using a code block having a vertical size of 64. In addition, when the image 2 in which the horizontal size of the input image is 4096 is encoded, the code block is encoded using the code block whose vertical size is 32.

  In this way, in the case of the image 2 that is relatively long in the horizontal direction, the vertical size of the code block is limited to 32, so that a large amount of line base is used compared to the case where the code block size of the vertical size 64 is used. Generation of the working memory can be suppressed. Since the working memory required at the time of processing does not change greatly during the processing of the image 1 and the processing of the image 2, the image can be stably encoded.

  According to the present embodiment, it is possible to suppress an increase in the amount of coefficient buffers processed in the memory due to a difference in image size, so that stable image coding is not affected by the difference in image size. Processing can be realized.

(Third embodiment)
In the third embodiment of the present invention, when determining the vertical size of the code block, if the horizontal size of the second input image is larger than the horizontal size of the first input image, the code block An image encoding apparatus will be described in which the size of the code block used for the second image is made smaller than the size of the code block used for the first image while maintaining the area.

  Also in this embodiment, a PC having the general hardware configuration described in the first embodiment is used (see FIG. 7). Moreover, it similarly has the function of the encoding process of the image encoding device shown in the first embodiment (see FIG. 8).

  The code block size affects the efficiency of entropy coding. This is because the context of entropy coding is initialized for each code block. Therefore, when determining the code block size, the number of coefficients included in the code block must not be too small. It has been empirically found that the size of 64 * 64 and the size of 32 * 32 are appropriate.

  For example, as shown in FIG. 17, consider a case where there are two types of input images and an image having a relatively short horizontal size and a long image are encoded. As shown in FIG. 17, an image 1 in which the horizontal size of the input image is 2048 pixels and an image 2 in which the horizontal size is 4096 pixels are encoded. When the image 1 is encoded, the code block size is 64 * 32 in length and becomes a vertically long rectangle with respect to the input direction. Also, when encoding image 2, the code block size is 32 * 64 in length and becomes a rectangle that is horizontally long in the input direction.

  When determining the vertical size of the code block, if the horizontal size of the second input image is larger than the horizontal size of the first input image, the code block area is kept constant, The vertical size of the code block used for the second image is made smaller than the vertical size of the code block used for the first image. That is, since the code block size is reduced while the code block area is kept constant, it means that the code block size is increased.

  Thus, by adjusting the vertical size while keeping the area of the code block constant, the code block can be kept constant without reducing the number of coefficients included in the code block, and an increase in memory consumption can be prevented. .

  The horizontal size of the input image used in the first to third embodiments may be the vertical size of the input image. In this case, the vertical size of the code block is the horizontal size of the code block. It is good also as a size. In other words, the horizontal direction of the input image may be replaced with the vertical direction of the input image, the vertical direction of the code block size may be replaced with the horizontal direction of the code block, and the horizontal direction of the code block size may be replaced with the vertical direction of the code block.

(Fourth embodiment)
In the above-described first to third embodiments, the form in which the entire subband is divided and encoded, that is, the form in which encoding is performed without mainly performing the precinct division has been described. In the fourth embodiment, an image encoding apparatus that encodes an image using precinct division will be described.

  Also in this embodiment, a PC having the general hardware configuration described in the first embodiment is used (see FIG. 7).

  FIG. 18 is a block diagram illustrating an encoding process of the image encoding device according to the present embodiment. According to FIG. 18, the image input unit 301 inputs an image read into the RAM 102 to be encoded as an input image. The precinct size determination unit 302 takes in the size of the image input in the image input unit 301 (the size in the horizontal direction of the image), and based on the image size, the precinct size determination unit 302 divides the precinct size in the subband coefficient space. Determine the size of the direction. The image encoding unit 303 inputs an image from the image input unit 301, performs wavelet transform processing on the input image using the precinct size determined by the precinct size determination unit 302, and encodes the input image. . Here, the encoding process performed in the image encoding unit 303 is a process for performing all encoding steps of quantization, coefficient bit modeling, and entropy encoding. The encoded image is temporarily written in the RAM 102 and written in a predetermined area on the HDD 103.

  As described above, the code block is cut out so as not to exceed the rectangular area of the precinct. That is, since the size of the code block does not exceed the size of the precinct, according to the present embodiment, it is possible to suppress an increase in memory consumption by changing the precinct size without changing the code block size. .

  For example, when the code block size is set to be a constant value of 64 * 64, and the horizontal size of the input image is large, for example, when precinct division is performed with a precinct size of 32 * 32, Since it is not necessary to hold subband coefficients exceeding the precinct size, the same effect of reducing the coefficient buffer amount as when the code block vertical size is reduced can be obtained.

  Here, when precinct division is performed, restrictions as shown in FIG. 19 are provided. According to FIG. 19, the precinct is divided by the magnitude of PPx power of 2 in the horizontal direction and PPy power of 2 in the vertical direction with (0, 0) as the origin. Also, the precinct division can determine the precinct size independently for each decomposition level. The same precinct size may be used for all decomposition levels, or only a specific decomposition level may be precinct divided.

  A subband having a resolution level of 0, that is, a subband having the maximum resolution has the largest number of coefficients included in one line, and therefore, a subband having a resolution level of 0 may be precinct divided by an appropriate size such as 32 * 32. It is most effective for saving memory.

  According to the present embodiment, it is possible to perform image encoding processing that suppresses an increase in memory consumption by determining the precinct size without changing the code block size while keeping the code block size constant. it can.

  As described above, the present invention has been described using the embodiment, but various modifications or improvements can be added to the above-described embodiment.

100 PC
101 CPU
102 RAM
103 HDD
104 System bus 201, 301 Image input unit 202 Code block size determining unit 203, 303 Image encoding unit 302 Precinct size determining unit

JP 2005-295540 A JP 2000-299863 A

ISO / IEC 15444-1, ITU-T TT. 800 JPEG 2000 Christos Chrysafis, Antonio Ortega, "Line Based, Reduced Memory, Wavelet Image Compression", IEEE Transactions on Image Processing, p 398-407.

Claims (11)

In an image encoding device that performs wavelet transform processing using a coefficient line buffer,
An image input means for inputting an image to be encoded;
Code block size determining means for determining a vertical size of a code block which is a minimum unit of encoding based on a horizontal size of an image input by the image input means;
Image encoding means for encoding an image using the code block size determined by the code block size determining means;
An image encoding device comprising:   The code block size determining means calculates the vertical size of the code block from the maximum capacity of the memory area available for wavelet transform processing using the horizontal size of the image input by the image input means. The image coding apparatus according to claim 1.   When the code block size determining unit determines the vertical size of the code block, the horizontal size of the second image is larger than the horizontal size of the first image input by the image input unit. 3. The image encoding according to claim 1, wherein the vertical size of the code block used for the second image is smaller than the vertical size of the code block used for the first image. apparatus.   When the code block size determining unit determines the vertical size of the code block, the horizontal size of the second image is larger than the horizontal size of the first image input by the image input unit. In this case, the vertical size of the code block used for the second image is made smaller than the vertical size of the code block used for the first image while maintaining the area of the code block. Item 4. The image encoding device according to any one of Items 1 to 3.   The code block size determination means uses a buffer having a number of lines equal to or larger than the vertical size of the code block in all subbands generated by the wavelet transform, and the horizontal direction of the image input by the image input means The image coding apparatus according to claim 1 or 2, wherein a size of the code block in a vertical direction is determined based on a size of the code block.   The code block size determination means uses a buffer having a number of lines equal to or larger than the vertical size of the code block in all subbands generated by the wavelet transform, and the horizontal direction of the image input by the image input means The image coding apparatus according to claim 1 or 2, wherein a size in the horizontal direction is determined simultaneously with a size in the vertical direction of the code block based on the size of the code block.   The image coding apparatus according to claim 1, wherein a size of the code block in a vertical direction is a value equal to or larger than a filter size used for the wavelet transform process.   8. The image code according to claim 1, wherein the horizontal size is replaced with a vertical size, and the vertical size is replaced with a horizontal size. Device. In an image encoding device that performs wavelet transform processing using a coefficient line buffer,
An image input means for inputting an image to be encoded;
Precinct size determining means for determining a vertical size of a precinct, which is a unit divided in a subband coefficient space, based on a horizontal size of an image input by the image input means;
Image encoding means for encoding an image using the size of the precinct determined by the precinct size determining means;
An image encoding device comprising:   The image encoding apparatus according to claim 1, wherein the image encoding method is a JPEG2000 method.   The image coding method performed in the image coding apparatus as described in any one of Claims 1-10.

Images