JP4036665B2 - Image encoding method and image decoding method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention belongs to the field of digital image processing, and relates to an image encoding method for encoding image data with high efficiency and an image decoding method for decoding encoded data encoded by this image encoding method.
[0002]
[Prior art]
As an image format for converting natural images into digital data and performing computer processing, a flash pix format (FlashPix Format Specification Version 1.0) has been proposed.
[0003]
In this format, data of a plurality of resolutions are simultaneously held in order to quickly extract data of a necessary resolution according to the capability of the display / printing apparatus and the user's request. In addition, the image is divided and held in units of tiles so that the load can be reduced by processing only necessary portions in the image data when the image is enlarged or reduced or edited.
[0004]
An encoding apparatus for encoding an image in accordance with the flash pix format will be described with reference to FIG. FIG. 32A is a diagram illustrating image reduction and tile division, and FIG. 32B is a block diagram illustrating an example of an encoding device.
[0005]
In the flash picks, first, 1/1 to 1/8 size images shown in images 1 to 4 in FIG. 32A are generated, and tile division and compression are performed on each image 1 to 4, respectively. There are features.
[0006]
First, the case where the image 1 of FIG. 32A is encoded by the encoding device of FIG. 32B will be described. Here, broken lines in images 1 to 4 in FIG. 32A represent tile boundaries.
[0007]
The original image is divided into tiles each having 64 pixels × 64 pixels by the tile dividing unit 3201, and then compressed for each tile by the JPEG compression unit 3202. The encoded data for each tile is integrated into one by the encoded data integration unit 3203 together with the tile division information from the tile division unit 3201, and the encoded data 1 is output.
[0008]
Next, the image 2 in FIG. 32A will be described. After the original image is reduced to 1/2 in both vertical and horizontal directions by the 1/2 reduction unit 3204, similarly, the encoded data 2 is obtained through the tile division unit 3205, the JPEG compression unit 3206, and the encoded data integration unit 3207.
[0009]
The reduction processing for generating the reduced image group (images 2 to 4) in FIG. 32A is repeated until the size of the entire reduced image is within one tile. In the example of FIG. 32A, the size of the image 3 does not fit in one tile, and is further reduced by ½ reduction processing when the size of the image 4 that fits in one tile is obtained. The process ends.
[0010]
The encoded data of image 3 is generated by a 1/2 reduction unit 3208, a tile division unit 3209, a JPEG compression unit 3210, and an encoded data integration unit 3211. The encoded data of image 4 is generated by a 1/2 reduction unit 3212, tile division. Generated by the unit 3213, the JPEG compression unit 3214, and the encoded data integration unit 3215.
[0011]
In this method, in addition to the encoded data of the 1/1 size image, the encoded data is also held for each of the reduced resolution images, so that the encoded data amount increases by about 1.4 times. On the other hand, at the time of encoding, a problem arises in that the amount of processing is large because compression processing is performed at each resolution.
[0012]
On the other hand, apart from flash pix, there is an image compression method by wavelet transform. In this method, image data with different resolution can be easily decoded from one encoded data that is compressed for the size of the original image. Therefore, there is no problem of an increase in the amount of encoded data due to support for multiple resolutions.
[0013]
That is, while the encoded data amount has increased by 1.4 times with the above-described flash pix, it is possible to respond to a request for decoding a plurality of resolutions with an encoded data amount of 1 ×.
[0014]
In the wavelet transform compression, the processing shown in the basic block diagram of FIG. 33 is performed. The original image becomes subband division data wavelet transformed by the wavelet transform unit 3301, quantized by the quantization unit 3302, entropy coded by the entropy coding unit 3303, and becomes coded data.
[0015]
FIG. 34 is a block diagram showing the wavelet transformation unit 3301 in FIG. 33 in more detail, and FIG. 35 shows image transformation by wavelet transformation. These are examples when three-dimensional two-dimensional subband division is performed.
[0016]
The original image in FIG. 35A is divided into two horizontal subbands by a horizontal low-pass filter 3401 and a horizontal high-pass filter 3402 in FIG. 34, and is divided by ½ subsampling units 3407 and 3408, respectively. Decimated by half.
[0017]
The divided two horizontal sub-bands are also sub-band divided by the low-pass filters 3403 and 3405 and the high-pass filters 3404 and 3406 and sub-sampling by the 1/2 sub-sampling units 3409 to 3412 in the vertical direction. At this point, it is converted into four subbands.
[0018]
Of these, the horizontal high band, the vertical high band subband (Fig. 34), the horizontal high band, the vertical low band (Fig. 34), the horizontal low band, the vertical high band. The subbands (H in FIG. 34) are wavelet transform coefficients indicated by H, R, and N in FIG. 35B, respectively.
[0019]
Subband division is repeated recursively only for the remaining low-frequency subband 3413 in both the horizontal and vertical directions.
[0020]
This recursive subband splitting includes horizontal low-pass filters 3414, 3426, horizontal high-pass filters 3415, 3427, vertical low-pass filters 3416, 3418, 3428, 3430, vertical high-pass filters 3417, 3419, 3429, 3431, and This is done by 1/2 sub-sampling units 3420-3425, 3432-3437.
[0021]
Note that the sub-bands (i) to (e) in FIG. 34 correspond to (i) to (b) in FIG.
[0022]
The wavelet transform coefficient of FIG. 35B obtained in this way is quantized by the quantization unit 3302 of FIG. 33 for each subband, and further entropy-encoded by the entropy encoding unit 3303 of FIG. Get. The entropy encoding unit 3303 can use Huffman encoding or arithmetic encoding.
[0023]
On the other hand, as shown in FIG. 36, wavelet transform decoding is performed by entropy decoding encoded data by an entropy decoding unit 3601, dequantizing by an inverse quantization unit 3602, and then subband synthesis by an inverse wavelet transform unit 3603. To obtain a decoded image.
[0024]
As a feature of encoding using wavelet transform, as shown in FIG. 35 (b), there is a point having a hierarchical structure corresponding to the resolution. For this reason, part or all of the encoded data is used for decoding. Thus, images with different resolutions can be easily decoded.
[0025]
That is, if the subbands a, b, c, and d in FIG. 35B are decoded, a quarter of the original image can be decoded, and in addition to this, h, f, and g can be decoded. For example, a 1/2 image can be decoded, and if all subbands are decoded, a 1/1 size image can be decoded.
[0026]
Here, the operation of the H-LP, H-HP, V-LP, and V-HP filters in FIG. 34 will be described with reference to FIG. FIG. 37 (b) is an enlarged view of a portion surrounded by a circle in FIG. 37 (a).
[0027]
In order to perform wavelet transform on the original image of FIG. 37A, when obtaining the output of a horizontal filter with a tap number of 9 bits for the pixel 3701 near the upper right end of the original image, the calculation target of the filter is indicated by 3702 Become an area.
[0028]
However, in this case, a part of the filter calculation target 3702 protrudes outside the original image, and no pixel data exists in this part. Similar problems arise with vertical filters.
[0029]
As described above, in the peripheral portion of the conversion target image, data outside the image is required according to the number of taps of the filter. If the subband division is repeated further, the region where the filter protrudes becomes wider.
[0030]
This problem is generally handled by a method such as folding an image at an end according to a certain rule.
[0031]
[Problems to be solved by the invention]
In the case of having separately encoded data for images of a plurality of resolutions such as flash pix, the load at the time of image data processing such as enlargement / reduction can be reduced, but the encoded data size is about 1.4. There is a disadvantage that doubles.
[0032]
On the other hand, when wavelet transform coding is used, a plurality of resolution data can be easily decoded from only one piece of coded data obtained by compressing the size of the original image, so that the coded data size does not increase.
[0033]
However, the method used in flash picks to divide an image into tiles and encode them in tile units (when a specific image area is the target of image processing, only the necessary image tiles are subject to image processing. If this is applied to the wavelet transform coding method, the filter used for the wavelet transform protrudes from the tile boundary.
[0034]
In other words, those using JPEG encoding such as Flash Picks are easy to encode in units of tiles because the encoding process is closed in the tile, whereas in wavelet transform encoding, the processing is easy. Since it protrudes around the tile, there is a problem that encoding processing and management in tile units becomes difficult.
[0035]
Furthermore, the conventional wavelet transform coding requires a memory that holds all the output of the wavelet transform unit 3301 in FIG. 33, that is, the wavelet transform coefficients in FIG. 35B. At this time, the wavelet transform coefficients are the same as the original image. Therefore, there is a problem that the required amount of memory becomes large. This problem becomes more prominent when dealing with high-resolution images.
[0036]
The present invention has been made in view of such a problem, and realizes high-function, high-efficiency coding with a small-scale hardware configuration by realizing the decoding of a plurality of resolutions and management by tile using wavelet transform. It is possible.
[0037]
[Means for Solving the Problems]
  The image encoding method of the present invention divides image data into tiles of N pixels × M pixels, and outputs N pixels × M pixels in the tile as encoding target data corresponding to each tile; The subband division is performed by extrapolating predetermined data around the encoding target data corresponding to the tile, and each tile is independently wavelet-encoded. Generating management information for decoding; and adding the management information to the encoded information to generate a bitstream, wherein the management information is a bitstream of encoded information of each tile. Information that indicates the position above and information that manages and identifies each subband.Therefore, they are arranged together at a position independent of the encoded information.It is characterized by that.
[0038]
  Image decoding of the present inventionMethodIs the image dataTheEach tile, and each tile is independentNiEncoded information obtained by wavelet encoding;Manage encoding informationA bitstream consisting of management information forAs an input, the decoded image corresponding to the required tile and resolutionAn image decoding method for decoding, comprising:The management information includes the size of the encoding information corresponding to each tile or resolution, and is collectively arranged at a position independent of the encoding information, and the storage position of the encoding information corresponding to the tile or resolution to be decoded is described above. Analyzing based on size, and based on the storage location,Performing wavelet decoding and said wavelet decodingTataIlunitConcatenated decoded imagesDoStep andAn image decoding method provided.
[0039]
DETAILED DESCRIPTION OF THE INVENTION
  Hereinafter, embodiments of the present invention will be described in detail. FIG. 1 shows an image encoding according to the first embodiment of the present invention.How to explainIt is a block diagram.
[0040]
The image data of the original image as shown in FIG. 2A is first divided into tiles of N pixels × M pixels determined in advance by the tile dividing unit 101. The divided image is shown in FIG. The tile dividing unit 101 outputs an image of N pixels × M pixels in the tile as data corresponding to each tile.
[0041]
Of the divided tiles, the subsequent processing will be described for the tile i in FIG. The wavelet transform unit 102 divides the image data of the tile i into subbands.
[0042]
Here, when subband division processing is performed near the periphery of the tile, data around the tile is extrapolated. That is, as shown in FIG. 37 (b), when the calculation target range 3702 of the filter used for the wavelet transformation protrudes outside the tile, data outside the tile is required. Insert and sub-band.
[0043]
As an extrapolation method, for example, as shown in FIG. 2C, a method of generating a mirror image by folding an image in a tile is used. Subsequently, the quantization unit 103 quantizes the wavelet transform coefficient, and the entropy encoding unit 104 performs entropy encoding to obtain encoded data of the tile i.
[0044]
For entropy coding, Huffman coding or arithmetic coding can be used. The wavelet transform unit 102, the quantization unit 103, and the entropy coding unit 104 are collectively referred to as a wavelet transform coding unit 105.
[0045]
On the other hand, the management information generation unit 106 uses the tile division information regarding the spatial position of each tile obtained from the tile division unit 101 and the information on each subband obtained from the wavelet transform coding unit 105, Management information for managing and identifying tiles and subbands is generated. This management information is used by the encoded data integration unit 107.
[0046]
The encoded data integration unit 107 uses the management information output from the management information generation unit 106 to organize and integrate the encoded information output from the entropy encoding unit 104, and to manage the management information in the bitstream. In addition, final encoded data is created.
[0047]
Here, encoded data is managed according to subbands and tiles when decoding an image, only images with different resolutions such as the example shown in FIG. 32A, or specific tiles in the image. This makes it possible to decrypt
[0048]
An example of the bit stream of the encoded data created in this way is shown in FIG. The bitstream is composed of a header that manages information of the entire bitstream and encoding information for each tile. The encoding information for each tile includes a tile header that manages information for each tile, and an image tile. It is composed of encoded information for each tile encoded by the wavelet transform encoding unit 105.
[0049]
In the tile header, information on bit positions corresponding to each subband is described. By referring to this, it is possible to know where a bit string corresponding to a necessary subband is located.
[0050]
Of course, the configuration of the bitstream according to the present invention is not limited to that shown in FIG. For example, the subband information of each tile is separated separately as shown in FIG. 4 (b) with respect to the one shown in FIG. 4 (a) having the same configuration as FIG. The tile information may be added to the band information to form an independent tile. In this way, it is possible to quickly reproduce the reduced overall image by accessing only the tiles of the reduced image.
[0051]
  Next, image coding according to Embodiment 2 of the present inventionMethodWill be described. Here, the image coding of the second embodimentMethodThe configuration is the same as the block diagram of the first embodiment described above with reference to FIG. 1, and only the operation of the tile dividing unit 101 is different. Therefore, hereinafter, the operation of the tile dividing unit 101 will be described with reference to FIG.
[0052]
In the tile dividing unit 101 of the first embodiment, after dividing an original image into tiles of N × M pixels, when outputting a specific tile to the wavelet transform unit 102, only the image data inside the tile is cut out as an output. However, the tile dividing unit 101 according to the second embodiment uses a unit that cuts out and outputs data by multiplying the original image by an appropriate window function.
[0053]
For example, when the tile ij in FIG. 5 is cut out, the result of multiplying the original image data by the window function FXi in the horizontal direction and then the window function FYj in the vertical direction is set as the output of the tile dividing unit 101. Note that i is a tile number in the horizontal direction, and j is a tile number in the vertical method.
[0054]
As a result, the result of multiplying the hatched image in FIG. 5 by the weight according to the window function is the output of the tile dividing unit 101. Here, as the window function, a window function having a total sum of 1 through all sections is used.
[0055]
That is,
ΣFXi (x) = 1 (0 ≦ x ≦ w)
ΣFYj (y) = 1 (0 ≦ y ≦ h)
Use a window function that satisfies
[0056]
However, w represents the width of the original image, h represents the height of the original image, and the x and y axes are assumed to have the upper left corner of the original image as the origin O, and are taken rightward and downward, respectively.
[0057]
Further, it is assumed that the sum of FXi (x) is taken for i and the sum of FXj (Y) is taken for j. FXi−1, FXi, FY1, FYj, and FYj + 1 in FIG. 5 represent a part of functions that satisfy such conditions.
[0058]
As a result of the data extraction by the window function, the output of the tile dividing unit 101 includes not only the pixels inside the tile ij but also surrounding pixels in the encoding target data with weights according to the value of the window function. become.
[0059]
  Next, the image coding of the first embodiment described aboveMethodImage decoding to decode data encoded withMethodWill be described as Embodiment 3 of the present invention. FIG. 6 shows image decoding according to the third embodiment.How to explainIt is a block diagram.
[0060]
  The input encoded data is the image encoding described in the first embodiment.MethodIt is encoded with. A management information separation unit 401 separates and extracts management information related to tile division and management information related to subbands from the encoded data.
[0061]
Based on the extracted management information, the encoded data extraction unit 402 determines and extracts encoded information portions of necessary tiles and subbands in the encoded information in response to a user request. In the example of the bit stream shown in FIG. 3, the management information is in the header and the tile header.
[0062]
The extracted encoded information is entropy-decoded by the entropy decoding unit 403 and dequantized by the inverse quantization unit 404 to obtain wavelet transform coefficients corresponding to the decoding target tile.
[0063]
The wavelet transform coefficient is subjected to inverse wavelet transform by the inverse wavelet transform unit 405, and a decoded image of the target tile is obtained. The entropy decoding unit 403, the inverse quantization unit 404, and the inverse wavelet transform unit 405 are collectively referred to as a wavelet transform decoding unit 406.
[0064]
Further, the tile connecting unit 407 connects the decoded tile groups based on the tile division information from the management information generating unit 401 to obtain a decoded image having a desired region / resolution.
[0065]
Referring to the example of the bitstream shown in FIG. 3, when decoding an entire image (all tiles) with a low resolution, a code corresponding to a subband with a low resolution while referring to the subband information of each tile header. .., Ia,..., Which are converted data parts, are wavelet transform decoded sequentially by the wavelet transform decoding unit 406 for each tile.
[0066]
Then, if the obtained low resolution tiles are connected by the tile connecting unit 407, an entire image with a low resolution can be obtained.
[0067]
In addition, when it is desired to enlarge a specific tile i from the low-resolution decoded image and display it at the highest resolution, the entire i-th tile encoded information that is encoded information corresponding to the tile i may be decoded.
[0068]
That is, if i-b is extracted in addition to the already extracted encoded information ia and decoded together with ia, a desired decoded image can be obtained. Of course, if all the encoded information (all tiles, all subbands) is decoded, it is possible to obtain decoded images of all regions with high resolution.
[0069]
As described above, an image of an arbitrary resolution and an arbitrary tile can be easily decoded according to a user request.
[0070]
  Next, image decoding according to the fourth embodiment of the present invention.MethodWill be described. The input encoded data is the image encoding described in the second embodiment.MethodIt is encoded by. Here, the image decoding of Embodiment 4MethodThe configuration is the same as that of the third embodiment described above with reference to FIG. 6, and only the operation of the tile connecting unit 407 is different. Therefore, hereinafter, the operation of the tile dividing unit 407 will be described with reference to FIG.
[0071]
  Image coding according to Embodiment 2MethodThen, since the pixel to be encoded of each tile includes the peripheral pixels of the tile, the size of the decoded data of the tile decoded by the wavelet transform decoding unit 406 is larger than the size of the tile.
[0072]
In FIG. 7, the tile is composed of 2 pixels × 2 pixels, and the size of the decoded data of the tile is 4 pixels × 4 pixels. In this case, the decoded data of the tile ij is a hatched portion in FIG. 7 and overlaps with the adjacent tile by the width of one pixel.
[0073]
The tile connecting unit 407 obtains a pixel value by adding the decoded data at the position where the decoded data overlaps when the tiles are connected. For example, for pixel a in FIG.
a (i-1, j-1) + a (i, j-1) + a (i-1, j) + a (i, j)
To calculate the pixel value.
[0074]
Here, a (i, j) represents the decoded data of the tile ij at the position of the pixel a.
[0075]
  Next, image coding according to Embodiment 5 of the present inventionMethodWill be described. FIG. 8 shows image encoding according to the fifth embodiment.How to explainIt is a block diagram.
[0076]
  Image coding of embodiment 5MethodIs the image coding of the first embodiment described above with reference to FIG.MethodThe difference is that when a tile is wavelet transform encoded, it is not extrapolated around the tile unconditionally, but another tile around the target tile is used if it exists. is there.
[0077]
As in the case of the first embodiment, as shown in FIG. 9A, the subsequent processing is described for the tile i in the original image divided by the tile dividing unit 501. When the image data of the tile i is converted by the wavelet transform unit 503, if a surrounding pixel exists in a region where the filter used for the wavelet transform protrudes from the tile i, the tile i is also wavelet transformed using the data of the pixel. .
[0078]
That is, in order to perform wavelet transform on the tile i in FIG. 9A, first, the wavelet indicated by hatching in FIG. 9B is selected from the tiles around the tile i in FIG. 9A. After the surrounding pixel area necessary for the conversion is added to the tile i, the wavelet conversion of the tile i is performed.
[0079]
This additional processing is performed by the surrounding pixel adding unit 502. Based on the tile division information obtained from the tile dividing unit 501, it is determined whether another tile exists around the encoding target tile, and the tile exists. In this case, necessary pixels are added.
[0080]
In the above example, the peripheral pixel adding unit 502 adds all the surrounding tiles and outputs tile image data. Therefore, the wavelet transform unit 503 to which the peripheral pixel adding unit 502 is input receives the wavelet according to the first embodiment that processes an image of a single tile. It is necessary to convert a large image as compared with the conversion unit 102.
[0081]
When the converted image becomes large, a device using the converted image requires a large work area, leading to an increase in cost and a decrease in operation speed. Therefore, another mode in which the converted image is made smaller is effective and will be described below.
[0082]
As shown in FIGS. 9C and 9D, the area added by the peripheral pixel adding unit 502 is limited to the x direction or the y direction, and the tile image data input to the wavelet transform unit 503 is reduced. It is.
[0083]
For example, in the case of FIG. 9C, necessary pixels are added when different tiles exist above and below the encoding target tile. For the left and right sides of the tile to be encoded, a method of generating a mirror image by folding the image in the tile is used. In the case of FIG. 9D, the top and bottom and the left and right are reversed from the case of FIG. 9C.
[0084]
As a method of performing wavelet transform, a method of repeating subband division using only one of FIGS. 9B, 9C, and 9D, or FIGS. 9B and 9C for each subband. ) And (d) are methods for switching the pixel addition method.
[0085]
Note that only the wavelet transform coefficient of the encoding target tile i is required as an output of the wavelet transform unit 503, and the pixels added by the surrounding pixel adding unit 502 are the wavelet transform coefficients of the pixels inside the tile i. Used only to calculate.
[0086]
Subsequently, quantization is performed by the quantization unit 504, and entropy coding is performed by the entropy coding unit 505, thereby obtaining coding information of the tile i. The wavelet transform unit 503, the quantization unit 504, and the entropy coding unit 505 are collectively referred to as a wavelet transform coding unit 506.
[0087]
On the other hand, the management information generation unit 507 uses the tile division information regarding the spatial position of each tile obtained from the tile division unit 501 and the information on each subband obtained from the wavelet transform coding unit 506, Management information for managing and identifying tiles and subbands is generated. This management information is used by the encoded data integration unit 508.
[0088]
The encoded data integration unit 508 uses the management information output from the management information generation unit 507 to organize and integrate the encoded information output from the entropy encoding unit 505, and to manage the management information in the bitstream. In addition, final encoded data is created as in the example shown in FIG. 3, for example.
[0089]
  Furthermore, the image coding according to the sixth embodiment of the present inventionMethodWill be described. Image coding according to Embodiment 6MethodThis configuration is the same as that of the fifth embodiment described above with reference to FIG. 8, and only the operation of the surrounding pixel adding unit 502 is different. Therefore, hereinafter, the operation of the surrounding pixel adding unit 502 will be described with reference to FIG.
[0090]
The processing of tile i in FIG. 10 will be described as an example. In the surrounding pixel adding unit 502 described as the fifth embodiment, when the tile i is input, all the pixels that are necessary for calculating the wavelet transform coefficient of the pixel in the tile i, that is, the pixels in the range where the filter protrudes are included in the tile i. It was added to. This range is a peripheral pixel range indicated by hatching in FIG.
[0091]
However, since the influence of a pixel far away from the tile i on the wavelet transform coefficient in the tile i is generally small, in the present embodiment, the result obtained by multiplying the peripheral pixel to be added by an appropriate weighting function is obtained. By adding to, the number of pixels to be added is reduced and the amount of calculation is reduced.
[0092]
As the weighting function, a function is used such that the portion close to the tile i is 1 and the distance is close to 0. The weighting function shown in FIG. 10 is an example. In the example of FIG. 10, as a result of multiplication by the weighting function, the pixels actually added are only effective pixel portions with halftone dots, and the outside is a pixel necessary for wavelet transform but is regarded as 0 and added. Not.
[0093]
As the weighting function, besides the one shown in FIG. 10, a step function that can be 1 if the distance from the tile i is within a certain standard and 0 if it is further away can be used.
[0094]
  Next, image coding according to Embodiment 7 of the present inventionMethodWill be described. FIG. 11 shows the image encoding according to the seventh embodiment.How to explainIt is a block diagram.
[0095]
  Image coding according to Embodiment 7MethodIs the image coding of the first embodiment described above with reference to FIG. 1 and the fifth embodiment described above with reference to FIG.MethodThe difference is that the wavelet transform unit 701 performs wavelet transform on the entire original image before tiling the original image, and then arranges the wavelet transform coefficients, which are the output of the wavelet transform unit 701, in tile units. It is a point which constitutes a tile instead.
[0096]
In FIG. 11, the original image is wavelet transformed by a wavelet transform unit 701 before being tiled. Next, the tile configuration unit 702 performs sorting to collect the wavelet transform coefficients corresponding to the same tile in space and configure the tiles.
[0097]
FIG. 12A shows an example of subbands obtained by wavelet transform by the wavelet transform unit 701. FIG. In this case, the coefficient b0 in the subband of the lowest frequency in FIG. 12A is the space between the coefficient parts b1, b2, b3, b4, b5, b6, b7, b8, b9 in the other subbands. In correspondence.
[0098]
Here, b1 to b3 are composed of 1 × 1, b4 to b6 are composed of 2 × 2, and b7 to b9 are composed of 4 × 4 coefficients. An embodiment in which these b0 to b9 are extracted from the respective sub-bands, and are configured in the form shown in FIG. 12B as one tile, and all other wavelet transform coefficients are rearranged in units of tiles. The result similar to that obtained when the original image is divided into tiles in step 5 and then wavelet transformed is obtained.
[0099]
Note that b0 does not have to be one coefficient, and may be a block of coefficients composed of k × 1 coefficients. In this case, b1 to b3 are composed of k × l, b4 to b6 are composed of 2k × 2l, and b7 to b9 are composed of 4k × 4l coefficients.
[0100]
The tiled wavelet transform coefficient output from the tile construction unit 702 is quantized by the quantization unit 703 and entropy-encoded by the entropy encoding unit 704 to become encoded information.
[0101]
On the other hand, the management information generation unit 706 uses the tile division information regarding the spatial position of each tile obtained from the tile configuration unit 702 and the information on each subband obtained from the wavelet transform coding unit 705. Management information for managing and identifying tiles and subbands is generated. This management information is used by the encoded data integration unit 707.
[0102]
The encoded data integration unit 707 uses the management information output from the management information generation unit 706 to organize and integrate the encoded information output from the entropy encoding unit 704, and the management information in the bitstream In addition, final encoded data is created as in the example shown in FIG. 3, for example.
[0103]
Note that the tile configuration unit 702 is arranged before the quantization unit 703, but the present invention is not limited to this. For example, the tile configuration unit 702 may be arranged after the quantization unit 703.
[0104]
  In addition, the image encoding according to any one of the fifth to seventh embodiments described above.MethodImage decoding to decode data encoded byMethodWill be described as Embodiment 8 of the present invention. FIG. 13 shows image decoding according to the eighth embodiment.How to explainIt is a block diagram. The encoded data to be input is the image encoding according to any one of the fifth to seventh embodiments.MethodIt is the encoded data encoded by.
[0105]
In FIG. 13, management information separation unit 901 separates management information related to tile division and management information related to subbands from the encoded data, and based on the extracted management information, the encoded data extraction unit 902 extracts the user information. In response to the request, a necessary encoded information part in the encoded information is determined and extracted. That is, encoded data corresponding to a necessary tile and resolution is extracted.
[0106]
The extracted encoded information is entropy-decoded by the entropy decoding unit 903 in units of tiles, dequantized by the inverse quantization unit 904, and wavelet transform coefficients corresponding to tiles necessary for decoding are obtained.
[0107]
The wavelet transform coefficient is subjected to inverse wavelet transform by the inverse wavelet transform unit 905, and a decoded image including data of surrounding pixels is obtained. The entropy decoding unit 903, the inverse quantization unit 904, and the inverse wavelet transform unit 905 are collectively referred to as a wavelet transform decoding unit 906.
[0108]
Further, the tile integration unit 907 integrates the decoded tile group based on the management information from the management information separation unit 901. Here, the spatially overlapped portion of the decoded image of each tile is superimposed to obtain the entire decoded image.
[0109]
  That is, in the second embodiment described above with reference to FIG. 5, wavelet transform is performed including the peripheral pixels of the tile. Also, the image coding according to the fifth embodimentMethodAs shown in FIG. 9B, the peripheral pixels of the tile are used at the time of wavelet transform. Similarly, the peripheral pixels are also used in the sixth embodiment described above with reference to FIG.
[0110]
  Also, the image coding according to the seventh embodimentMethodHowever, the process using the peripheral pixels of the tile is not clearly described, but when the entire original image is wavelet transformed, the process equivalent to the fifth embodiment is performed in principle.
[0111]
For this reason, when wavelet transform decoding is performed by the wavelet transform decoding unit 906 in FIG. 13, peripheral pixel data is generated, and the tile integration unit 907 superimposes the peripheral pixels of the decoded tile on adjacent tiles. For the superposition, addition between pixels is used.
[0112]
  Next, image decoding according to Embodiment 9 of the present inventionMethodWill be described. This is the image decoding of the eighth embodiment.MethodSimilarly to the image coding according to any one of Embodiments 5 to 7MethodDecoding using encoded data encoded withMethodIt is. FIG. 14 shows image decoding according to the ninth embodiment.How to explainIt is a block diagram.
[0113]
In FIG. 14, the management information separation unit 1001 separates and extracts management information related to tile division and management information related to subbands from the encoded data, and the encoded data extraction unit 1002 extracts the management information based on the extracted management information. In accordance with a user request, a necessary encoded data portion in the encoded information is determined and extracted. That is, encoding information corresponding to a necessary tile and resolution is extracted.
[0114]
The extracted encoding information is entropy-decoded by the entropy decoding unit 1003 in units of tiles, and dequantized by the inverse quantization unit 1004, and wavelet transform coefficients corresponding to tiles necessary for decoding are obtained. Here, the wavelet transform coefficient rearrangement unit 1005 rearranges the wavelet transform coefficients to the state before tiling.
[0115]
That is, the wavelet transform coefficients divided into tile units shown in FIG. 12B are rearranged into the state shown in FIG. When all the tiles have been processed, the entire wavelet transform coefficient shown in FIG. 12A is obtained.
[0116]
Since the rearranged wavelet transform coefficients can be decoded by a single inverse wavelet transform, if the wavelet transform coefficients are inversely wavelet transformed by the inverse wavelet transform unit 1006, the entire decoded image can be obtained.
[0117]
The entropy decoding unit 1003, the inverse quantization unit 1004, and the inverse wavelet transform unit 1006 are collectively referred to as a wavelet transform decoding unit 1007. The wavelet transform coefficient rearranging unit 1005 is arranged at the subsequent stage of the inverse quantization unit 1004, but is not limited to this, and may be disposed, for example, before the inverse quantization unit 1004.
[0118]
  Next, image coding according to Embodiment 10 of the present inventionMethodWill be described. FIG. 15 (e) shows the image coding of the first embodiment, the second embodiment, the fifth embodiment, and the sixth embodiment.Method10 is a block diagram showing a portion corresponding to a wavelet transform unit (102 in FIG. 1, 503 in FIG. 8).
[0119]
A memory 1102 in FIG. 15E is for storing the wavelet transform coefficients that are sub-band divided by the wavelet transform unit 1101. At this time, only the wavelet transform coefficient corresponding to the tile currently being processed by the wavelet transform unit 1101 is stored in the memory 1102, and when the wavelet transform of the tile is completed, the data is converted into a quantization unit (FIG. 1) as the next step. 103 and 504) of FIG.
[0120]
Therefore, the amount of data to be stored in the memory 1102 does not correspond to the entire image, and can be suppressed to the amount of data necessary to wavelet transform one tile.
[0121]
That is, in the wavelet transform without tiling, as shown in FIG. 15A, the conversion target is the entire image, and all the wavelet transform coefficients of FIG. 15B, which is the output of the wavelet transform unit 1101, are stored in the memory. For example, as shown in FIG. 15C, it is necessary to prepare only a memory that can store the wavelet transform coefficients corresponding to FIG. 15D by performing tiling as shown in FIG. Thus, the required memory amount can be greatly reduced.
[0122]
  Image decodingMethodBut the same effect can be expected. Image decoding according to the eleventh embodiment of the present inventionMethodWill be described. FIG. 16E illustrates the image decoding described above as the third, fourth, and eighth embodiments.MethodIt is the block diagram which showed the part corresponding to an inverse wavelet transformation part (405 of FIG. 6, 905 of FIG. 13) among these.
[0123]
First, wavelet transform coefficients necessary to decode one tile are stored in the memory 1201 in FIG. 16E, and subband synthesis is performed by the inverse wavelet transform unit 1202.
[0124]
Therefore, when the decoding target image is shown in FIG. 16B, in the wavelet transform that is not tiled, the amount of data to be stored in the memory is all the wavelet transform coefficients shown in FIG. As shown in FIG. 16D, when decoding a tiled image, the amount of data to be stored in the memory 1201 of this embodiment is the wavelet transform coefficient corresponding to FIG. The amount of memory is greatly reduced.
[0125]
As described above, any of the embodiments of the present invention described above can be configured by adaptively switching using a plurality of subband division filters at the time of wavelet transform in encoding.
[0126]
Here, the sub-band division filter is a low-pass filter and a high-pass filter used for the sub-band division described as the conventional example. In the wavelet transform, subband division is repeated. At this time, there are various types of filters used in each subband division depending on the number of taps and coefficient values.
[0127]
Therefore, if an appropriate filter is used for each subband division, the necessary amount of peripheral pixels of the image to be encoded required by the wavelet transform coefficient can be controlled for each subband, and the balance between the processing amount and the image quality can be controlled. The optimal wavelet transform can be performed.
[0128]
  Such image codingMethodDecoding corresponding toMethodThen, using a subband synthesis filter corresponding to the subband division filter used at the time of wavelet transformation, the inverse wavelet transformation is performed while switching the filters in each subband synthesis.
[0129]
  Next, image coding according to the twelfth embodiment of the present inventionMethodWill be described. In the present embodiment, the input image can be encoded by one of a plurality of predetermined encoding methods.
[0130]
  FIG. 17 shows image encoding according to the twelfth embodiment.MethodExampleTo explainIt is a block diagram, and in the present embodiment, encoding is performed by switching between the system of the first embodiment and the system of the seventh embodiment.
[0131]
In FIG. 17, a tile wavelet encoding unit 1601 performs wavelet encoding on an input image in units of tiles, and outputs encoding information. Further, the tile wavelet encoding unit 1601 outputs tile division information, subband information, and flag information.
[0132]
The management information generation unit 1603 receives the tile division information, the subband information, and the flag information as inputs, and generates and outputs management information by combining them. The encoded data combining unit 107 outputs encoded data obtained by adding the encoded information and management information.
[0133]
In the tile wavelet encoding unit 1601, the input original image is divided by the tile dividing unit 101, and the divided image is input to the terminal 0 of the first switch 1604. Further, the original image is input as it is to the terminal 1 of the first switch 1601. One of these outputs is input to the wavelet encoding unit 1607 via the first switch 1604.
[0134]
The wavelet encoding unit 1607 performs wavelet encoding on the input image. The output of the first wavelet transform unit 1608 is directly input to the quantization unit 103 via the second switch 1605 or is further input to the quantization unit 103 via the tile configuration unit 702.
[0135]
The operation of the first wavelet transform unit 1608 is the same as that of the wavelet transform unit 102 in the first embodiment described above with reference to FIG.
[0136]
Then, the flag generation unit 1602 outputs a flag indicating whether the encoding system of the first embodiment or the encoding system of the seventh embodiment is used, and at the same time, the first switch 1604, the second switch 1605, and the third switch 1606 is controlled.
[0137]
If each switch 1604, 1605, 1606 is coupled to the terminal 0, the same processing as that of the encoding in the first embodiment is performed, and if it is coupled to the terminal 1, the encoding in the seventh embodiment is performed. Performs the same processing as
[0138]
The operation of the tile configuring unit 702 is the same as that of the seventh embodiment described above with reference to FIG.
[0139]
As described above, according to the present embodiment, encoding can be performed in units of tiles, and encoding is performed for each image by the method of the first embodiment that is easy to process, or the processing is slightly complicated. It is possible to selectively switch whether encoding is performed according to the method of the seventh embodiment in which no distortion occurs at the tile boundary.
[0140]
  FIG. 18 shows image encoding according to the twelfth embodiment.MethodAnother example ofTo explainFIG. 4 is a block diagram, and in the present example, encoding can be performed by switching between the system of the first embodiment and the system of the fifth embodiment.
[0141]
  Image coding of this embodimentMethod18, the tile configuration unit 702 related to the seventh embodiment in FIG. 17 is deleted, the peripheral pixel adding unit 502 and the second wavelet encoding unit 1705 related to the fifth embodiment are added, and The switch for switching between these has been changed. Operations other than the tile wavelet encoding unit 1701 and the wavelet encoding unit 1702 in FIG.
[0142]
The wavelet encoding unit 1702 performs wavelet encoding on the input image and outputs encoding information. There are two types of inputs, one connected to the first wavelet transform unit 1608 and the other connected to the second wavelet transform unit 1705.
[0143]
When an image is input to the first wavelet transform unit 1608, the wavelet transform unit 1702 performs the same operation as the wavelet encoding unit 1607. On the other hand, when the image is input to the second wavelet transform unit 1705, the processing of the second wavelet transform unit 1705 is the same as that of the wavelet transform unit 503, and therefore the wavelet encoding unit 1702 is the wavelet encoding unit 506. Behaves the same as
[0144]
In the tile wavelet encoding unit 1701, the input image is divided into tile units and input to the first switch 1703. On the other hand, the peripheral image is added to the divided image and input to the second switch 1704. The flag generation unit 1706 selects whether the wavelet encoding unit 1702 uses the first wavelet transform unit 1608 or the second wavelet transform unit 1705, and outputs a flag indicating this.
[0145]
At the same time, control is performed such that only one of the first switch 1703 and the second switch 1704 is turned on. That is, when the first switch 1703 is on, the divided image is input to the first wavelet transform unit 1608, and processing equivalent to that encoded by the method of the first embodiment is performed. When the second switch 1704 is on, the divided image and the surrounding image are input to the second wavelet transform unit 1705, and processing equivalent to that performed by the method of the fifth embodiment is performed.
[0146]
As a result, encoding can be performed on a tile-by-tile basis, and encoding can be performed for each image by the method of the first embodiment that is easy to process, or the processing is slightly complicated, but no distortion occurs at the tile boundary. It is possible to selectively switch the encoding according to the method of the fifth mode.
[0147]
  Further, FIG. 19 shows the image encoding of the twelfth embodiment.MethodAnother example ofTo explainFIG. 4 is a block diagram, and in the present example, encoding can be performed by switching between the scheme of the first embodiment, the scheme of the fifth embodiment, and the scheme of the seventh embodiment.
[0148]
  Image coding of this embodimentMethodAs shown in FIG. 19, a tile configuration unit 702 according to the seventh embodiment is added in FIG. 18, and a switch for switching these is changed. Operations other than the tile wavelet encoding unit 1801 and the wavelet encoding unit 1807 in FIG.
[0149]
A wavelet encoding unit 1807 performs wavelet encoding of the input image and outputs encoding information. The output of the first wavelet transform unit 1608 is directly input to the quantization unit 103 via the third switch 1805 or is further input to the quantization unit 103 via the tile configuration unit 702. The output of the second wavelet transform unit 1705 is directly input to the quantization unit 103.
[0150]
In the tile wavelet encoding unit 1801, the input image is directly input to the terminal 0 of the first switch 1803, or is input to the terminal 1 of the first switch 1803 after being divided into tiles, or the divided image is divided. An image in which the surrounding pixels are added to the tile is input to the terminal 2 of the first switch 1803.
[0151]
These images are input to the first wavelet transform unit 1608 or the second wavelet transform unit 1705 via the second switch 1804, and output as encoded information via the quantizer 103 and the entropy encoder 104. The
[0152]
The flag generating unit 1802 controls the first switch 1803, the second switch 1804, the third switch 1805, and the fourth switch 1806, and switches the three modes 0, 1, and 2. The numbers shown on the terminals of the respective switches 1803, 1804, 1805, and 1806 indicate the mode numbers.
[0153]
For example, when the first switch 1803 is connected to the terminal 0, the remaining switches 1804, 1805, and 1806 are also connected to the terminal 0. For this reason, when each of the switches 1803, 1804, 1805, and 1806 is connected to the terminal 0, processing equivalent to that encoded by the method of the seventh embodiment is performed.
[0154]
When each of the switches 1803, 1804, 1805, and 1806 is connected to the terminal 1, the same processing as that performed in the method of the first embodiment is performed, and the first switch 1803, the second switch 1804, the fourth switch When the switch 1806 is connected to the terminal 2, the same processing as that performed by the method of the fifth embodiment is performed.
[0155]
As a result, encoding can be performed on a tile-by-tile basis, and encoding can be performed for each image by the method of the first embodiment that is easy to process, or the processing is slightly complicated, but no distortion occurs at the tile boundary. It is possible to selectively switch whether encoding is performed according to the scheme of the fifth or seventh embodiment.
[0156]
  Next, image decoding according to the thirteenth embodiment of the present invention.MethodWill be described. This is the image encoding described above as the twelfth embodiment.MethodImage decoding to decode data encoded withMethodIt is. In the present embodiment, input encoded data is decoded by selecting one of a plurality of predetermined decoding methods.
[0157]
  FIG. 20 illustrates image decoding according to the thirteenth embodiment.MethodExampleTo explainIt is a block diagram and the image decoding of a present ExampleMethodIn the method, encoded data encoded by switching between the method of the first embodiment and the method of the seventh embodiment can be decoded.
[0158]
In FIG. 20, the encoded information and the management information separated by the management information separation unit 401 are input to the tile wavelet decoding unit 1901, respectively. The tile wavelet decoding unit 1901 performs decoding in tile units using the encoded information and management information, and outputs a decoded image.
[0159]
The encoded information is input to the wavelet decoding unit 1902 and subjected to wavelet decoding. The image decoded by the wavelet decoding unit 1902 is directly output via the second switch 1904 or further output via the tile connecting unit 407.
[0160]
In the wavelet decoding unit 1902, the output of the inverse quantization unit 404 is input directly to the first inverse wavelet transform unit 1906 via the first switch 1903 or further through the wavelet coefficient rearrangement unit 1005. To the inverse wavelet transform unit.
[0161]
The operation of the first inverse wavelet transform unit 1906 is the same as that of the inverse wavelet transform unit 405 in the third embodiment described above with reference to FIG.
[0162]
  The flag extraction unit 1905 extracts a flag for controlling the first switch 1903 and the second switch 1904 from the management information. When the switches 1903 and 1904 are connected to the terminal 0, the image decoding according to the third embodiment is performed.MethodWhen the same operation is performed and the terminal 1 is connected, the image decoding according to the ninth embodiment is performed.MethodPerforms the same operation as.
[0163]
The operation of the tile configuring unit 407 is the same as that of the third embodiment described above with reference to FIG.
[0164]
As described above, according to the present embodiment, decoding can be performed in units of tiles, and decoding can be performed for each image by the method of the third embodiment in which processing is simple or processing is slightly complicated. It is possible to selectively switch whether decoding is performed according to the method of the ninth embodiment in which no distortion occurs.
[0165]
  FIG. 21 shows image decoding according to the thirteenth embodiment.MethodAnother example ofTo explainIt is a block diagram and the image decoding of a present ExampleMethodThe encoded data encoded by switching between the method of the first embodiment and the method of the fifth embodiment can be decoded.
[0166]
In FIG. 21, the operations of the parts other than the tile wavelet decoding unit 2001 and the wavelet decoding unit 2002 are the same as those in FIG.
[0167]
The wavelet decoding unit 2002 performs wavelet decoding on the input encoded information. At this time, the output of the inverse quantization unit 404 is input to the first inverse wavelet transform unit 1906 or the second inverse wavelet transform unit 2003 via the first switch 2004.
[0168]
The output of the inverse first wavelet transform unit 1906 is input to the tile connection unit 407, and the output of the second wavelet transform unit 2003 is input to the tile integration unit 907.
[0169]
The operation of the second inverse wavelet transform unit 2003 is the same as that of the inverse wavelet transform unit 905 in the eighth embodiment described above with reference to FIG.
[0170]
The tile wavelet decoding unit 2001 performs wavelet decoding on the encoded information input by the wavelet decoding unit 2002, and the output of the wavelet decoding unit 2002 is connected to either the tile connection unit 407 or the tile integration unit 907, and the decoded image Is played.
[0171]
  On the other hand, the flag extraction unit 2005 extracts a flag from the input management information, and the first switch 2004 is switched by the extracted flag. When the first switch 2004 is connected to the terminal 0, the image decoding according to the third embodiment is performed.MethodWhen the same operation is performed and the terminal 1 is connected, the image decoding according to the eighth embodiment is performed.MethodBehaves the same as
[0172]
As a result, decoding can be performed in units of tiles, and decoding can be performed for each image by the method of the third embodiment in which processing is simple, or the processing is slightly complicated, but distortion in the tile boundary does not occur. It is possible to selectively switch whether the decoding is performed by the method.
[0173]
  Furthermore, FIG. 22 shows image decoding according to the thirteenth embodiment.MethodAnother example ofTo explainIt is a block diagram and the image decoding of a present ExampleMethodIn the method, encoded data encoded by switching the method of the first embodiment, the method of the fifth embodiment, and the method of the seventh embodiment can be decoded.
[0174]
  Image decoding of this embodimentMethodAs shown in FIG. 22, in FIG. 21, a wavelet coefficient rearranging unit 1005 is added, and a switch for switching them is changed. In the figure, the operations of the parts other than the tile wavelet decoding unit 2101 and the wavelet decoding unit 2102 are the same as those in FIG.
[0175]
The wavelet decoding unit 2102 performs wavelet decoding on the input encoded information. At this time, the output of the inverse quantization unit 404 is directly input to the first inverse wavelet transform unit 1906 via the terminal 0 of the first switch 2103, or the terminal 1 of the first switch 2103 and the wavelet coefficient rearrangement unit The first inverse wavelet transform unit 1906 or the second inverse wavelet transform unit 2003 via the terminal 2 of the first switch 2103.
[0176]
The output of the first inverse wavelet transform unit 1906 is input to the tile connecting unit 407 via the second switch 2104 or a decoded image is directly output. The output of the second inverse wavelet transform unit 2003 is input to the tile integration unit 907. Since the operation of other parts is the same as that of the wavelet decoding unit 2002, the description thereof is omitted.
[0177]
In the tile wavelet decoding unit 2101, the flag extraction unit 2105 extracts a flag from the management information. The first switch 2103 and the second switch 2104 are controlled by the extracted flag information. The remaining management information is input to the tile connecting unit 407 and the tile integrating unit 907.
[0178]
  When the switches 2103 and 2104 are connected to the terminal 0, the image decoding according to the third embodiment is performed.MethodWhen the same operation is performed and the terminal 1 is connected, the image decoding according to the ninth embodiment is performed.MethodWhen the first switch 2103 is connected to the terminal 2, the image decoding according to the eighth embodiment is performed regardless of the connection destination of the second switch 2104.MethodPerforms the same operation as.
[0179]
In this way, decoding can be performed in units of tiles, and encoding is performed for each image by the method of the third embodiment, which is easy to process, or the processing is slightly complicated, but distortion does not occur at the tile boundary. Alternatively, it is possible to selectively switch whether the decoding is performed according to the method of the ninth embodiment.
[0180]
  Next, image coding according to Embodiment 14 of the present inventionMethodWill be described. In the present embodiment, information for distinguishing tiles is added to management information for managing tiles so that encoding information of a target tile can be decoded at high speed.
[0181]
  FIG. 23 shows image coding according to the fourteenth embodiment.MethodExampleTo explainIt is a block diagram. In FIG. 23, an input original image is encoded in tile units by a tile wavelet encoding unit 2201, and information for management (for example, tile division information, flag information, subband information) and encoding information are included. Generated.
[0182]
The ID generation unit 2202 generates ID information for distinguishing each tile. The management information generation unit 2203 adds management information and the ID information to generate management information. The encoded data combining unit 2204 combines the encoded information and the management information, and further adds a start code indicating the head of the tile to the head of each tile to generate encoded data.
[0183]
  As an example of the format of the encoded data, as shown in FIG. 24A, the information of each tile is composed of the start code of the tile, management information (tile header), and encoded information. The tile wavelet encoding unit 2201 performs image encoding in the first embodiment, the second embodiment, the fifth embodiment, the sixth embodiment, the seventh embodiment, the tenth embodiment, the twelfth embodiment, and the fourteenth embodiment.MethodCan be used.
[0184]
Here, in order to distinguish the tiles obtained by dividing the original image, 1, 2,. . . And ID information are assigned, tiles can be encoded in an arbitrary order, and the order can be changed after encoding. If the order of encoding the tiles is determined in advance, the ID generation unit 2202 can be omitted.
[0185]
Since each tile starts with a start code, it is possible to identify where each tile is located using this as a mark. Instead, when the data amount of the tile (a combination of the encoding information and the tile header) is used, it is possible to identify where each tile is located.
[0186]
  FIG. 25 shows image encoding according to the fourteenth embodiment.MethodAnother example ofTo explainFIG. 24 is a block diagram showing the image encoding shown in FIG.MethodIn addition, a data amount measuring unit 2301 for calculating the size of the tile is added, and description of operations other than the data amount measuring unit 2301 and the management information generating unit 2302 is omitted.
[0187]
In FIG. 25, the data amount measuring unit 2301 measures the amount of data encoded for each tile and outputs this. The management information generation unit 2302 generates management information by adding the information for management, ID information, and the amount of tile data.
[0188]
As an example of the format of the encoded data, as shown in FIG. 24B, the data amount of the encoded information of the tile is arranged at the head of each tile, and subsequently, the other management information (tile header) and the code Followed by conversion information. Note that the data amount of tiles does not necessarily have to be arranged at the head of each tile, and can be collected at the head, for example.
[0189]
  Further, FIG. 26 shows the image encoding of the fourteenth embodiment.MethodAnother example ofTo explainFIG. 26 is a block diagram showing the image encoding shown in FIG.MethodThe encoded data rearranging unit 2401 is added to the above, and the description of the operation of other parts is omitted.
[0190]
In FIG. 26, the encoded data rearranging unit 2401 extracts the data amount of each tile from the encoded data created by the encoded data combining unit 2204, arranges these at the head of the encoded data, and then performs the rest. Arrange in order and output encoded data.
[0191]
As an example of the format of the encoded data, as shown in FIG. 24C, the position of the target tile can be easily calculated by adding the data amounts of all the tiles arranged at the head. .
[0192]
  The same effect can be obtained with the configuration shown in FIG. FIG. 27 shows image encoding according to the fourteenth embodiment.MethodAnother example ofTo explainFIG. 26 is a block diagram showing the image encoding shown in FIG.MethodIn addition, an encoded data storage buffer 2501 and a management information storage buffer 2502 are added, and description of operations other than the encoded data storage buffer 2501, the management information storage buffer 2502, and the encoded data combining unit 2503 is omitted.
[0193]
In FIG. 27, the encoded information output from the tile wavelet encoding unit 2201 is temporarily stored in the encoded data storage buffer 2501. The management information accumulation buffer 2502 accumulates the management information of each tile generated by the management information generation unit 2302, extracts the tile data amount from the management information, and outputs this to the encoded data combination unit 2503. Next, the remaining management information is output.
[0194]
The encoded data combining unit 2503 first outputs the data amount of all the input tiles, and combines and outputs the remaining management information and encoded information.
[0195]
As described above, according to the present embodiment, it is possible to search and decode the encoded information of the tile to be decoded from the encoded data at high speed.
[0196]
  Next, image decoding according to the fifteenth embodiment of the present invention.MethodWill be described. FIG. 28 shows image decoding according to the fifteenth embodiment.MethodTheTo explainIt is a block diagram, and the present embodiment is the image coding of the fourteenth embodiment described aboveMethodImage decoding to decode data encoded withMethodIt is.
[0197]
In FIG. 28, the decoding tile determination unit 2603 determines the tile ID to be decoded according to the user's instruction. The management information separation unit 2606 retrieves a start code indicating the head of each tile from the encoded data, and separates the management information and the encoded information regarding the tile.
[0198]
Based on the management information, the data skip control unit 2602 determines whether the tile ID of a tile to be decoded is the determined tile ID. If this tile ID is the tile ID, the first switch 2605 and the second switch Turn on 2604. In this way, the tile wavelet decoding unit 2601 can decode only a specific tile.
[0199]
  When the data amount of the tile is described in the tile management information, the management information separation unit 2606 does not need to search the head of each tile, and it is only necessary to skip the amount of data described. The tile wavelet decoding unit 2601 performs image decoding according to the third, fourth, eighth, ninth, eleventh, eleventh, thirteenth, and fifteenth embodiments.MethodCan be used.
[0200]
As described above, according to the present embodiment, it is possible to quickly decode a target tile by decoding only the management information at the head of the tile without decoding all the encoded data.
[0201]
  Next, image coding according to Embodiment 16 of the present inventionMethodWill be described. In the present embodiment, peripheral tile information is also added to management information for managing tiles, and encoded information of peripheral tiles can be decoded at high speed.
[0202]
  FIG. 29A shows the image encoding according to the sixteenth embodiment.MethodExampleTo explainIt is a block diagram. Image coding of this embodimentMethodFIG. 23 is obtained by adding a peripheral tile ID determination unit 2801 to the fourteenth embodiment shown in FIG. 23, and the operation of the management information generation unit 2802 is different. For this reason, description of parts other than the peripheral tile ID determination unit 2801 and the management information generation unit 2802 is omitted.
[0203]
  Note that the tile wavelet encoding unit 2801 performs image encoding according to the fifth, sixth, seventh, tenth, twelfth, and fourteenth embodiments.MethodCan be used.
[0204]
In FIG. 29A, the peripheral tile ID determination unit 2801 determines peripheral tile IDs necessary for decoding from tile division information, flag information, subband information, and tile IDs generated by the ID generation unit 2202. The management information creation unit 2802 generates management information in which tile division information, flag information, subband information, and tile ID are added to the surrounding tile ID.
[0205]
Note that the plurality of tile IDs determined by the peripheral tile ID determination unit 2801 do not have to be all tile IDs necessary for encoding. For example, as shown in FIG. You may limit to tile ID of the tile located in the upper left and lower left.
[0206]
As an example of the format of the encoded data, a configuration in which the management information (tile header) includes a tile ID and peripheral tile IDs in FIG.
[0207]
  FIG. 30 shows the image encoding according to the sixteenth embodiment.MethodAnother example ofTo explainIt is a block diagram, and seeks to speed up the search for encoded information tiled at the time of decoding by including position information of neighboring tiles in management information. Image coding of this embodimentMethodIn FIG. 27, the management information storage buffer 2502 is deleted from the fourteenth embodiment shown in FIG. 27, and a data amount storage unit 2901, a relative position calculation unit 2902, and an information storage buffer 2904 are added.
[0208]
Since the operations other than the data amount storage unit 2901, the relative position calculation unit 2902, the information accumulation buffer 2904, the management information generation unit 2903, and the ID generation unit 2905 are the same as those described above, description thereof will be omitted.
[0209]
In FIG. 30, all the encoded information output from the tile wavelet encoding unit 2201 is stored in the encoded data storage buffer 2501, and the tile division information, flag information, and sub information output from the tile wavelet encoding unit 2201 are stored. All pieces of band information are stored in the information storage buffer 2904. The data amount of the encoded information of each tile output from the data amount measuring unit 2301 is all stored in the data amount storage unit 2901.
[0210]
The ID generation unit 2905 outputs ID information for distinguishing each tile, and outputs information stored in the information storage buffer 2904, the data amount storage unit 2901, and the encoded data storage buffer 2501 in units of tiles. Control as follows. The data amount storage unit 2901 outputs the data amount of the tile to the management information generation unit 2903 based on the input tile ID, and is necessary for calculating the relative position of the tile having the tile ID and the surrounding tile. The data amount of the tile is output to the relative position calculation unit 2902.
[0211]
The relative position calculation unit 2902 calculates the relative position where the encoding information of the neighboring tiles exists with respect to the tile to be encoded, using the input data amount of each tile, and outputs the result. The management information generation unit 2903 generates management information from the input tile ID information, tile division information, flag information, subband information, tile data amount, relative position of the surrounding tiles, etc. Output.
[0212]
In this way, by decoding only the management information at the head of the tile without decoding all the encoded data, the encoded data can be generated so that the target tile and surrounding tiles necessary for decoding can be quickly decoded. It becomes possible to do.
[0213]
  Next, the image decoding according to the nineteenth embodiment of the present invention.MethodWill be described. FIG. 31 shows image decoding according to the nineteenth embodiment.MethodTheTo explainIt is a block diagram, and the present embodiment is the image coding according to the eighteenth embodiment described above.MethodImage decoding to decode data encoded withMethodIt is.
[0214]
In this embodiment, a buffer 3001 is added to the fifteenth embodiment shown in FIG. 28, and the operations other than the buffer 3001 and the data read skip controller 3002 are the same as those in FIG. Omitted.
[0215]
In FIG. 31, input encoded data is stored in a temporary buffer 3001 and sequentially output. The data skip control unit 3002 extracts the ID of the tile to be decoded based on the input management information, and if this is the determined tile ID or the tile ID of the surrounding tile, the first switch 2605 and the second switch Switch 2604 is turned on.
[0216]
If the management information includes tile IDs of peripheral tiles necessary for decoding, control is performed so that encoded information of the peripheral tiles is output from the buffer 3001. In this way, the tile wavelet decoding unit 2601 can decode a specific tile and its surroundings.
[0217]
Here, when the decoded peripheral tile ID included in the management information is a predetermined number smaller than the number of peripheral tiles (for example, tiles indicated by halftone dots in FIG. 29B), it is necessary for decoding. Tile IDs at other positions (white tiles in FIG. 29B) are determined from the decoded peripheral tile IDs.
[0218]
  The tile wavelet decoding unit 2601 performs image decoding according to the eighth, ninth, eleventh, thirteenth, and fifteenth embodiments.MethodCan be used.
[0219]
Thus, by decoding only the management information at the head of the tile without decoding all the encoded data, it becomes possible to quickly decode the target tile and surrounding tiles necessary for decoding.
[0220]
  As described above, the image encoding of this embodimentMethodAnd image decodingMethodBy using, it becomes possible to easily decode a decoded image having a resolution according to a user's request without increasing the amount of encoded data. This is a significant advantage compared to an increase in the amount of encoded data by 1.4 times because the flash pix using JPEG supports multiple resolutions.
[0221]
In addition, when dividing an image into tiles and enabling decoding of only a specific area, the encoding by wavelet transform is difficult in principle to be closed in the tile, and is unsuitable for tile division processing. On the other hand, in the present invention, encoding / decoding processing in units of tiles is possible using wavelet transform.
[0222]
In other words, by encoding the image in units of tiles, if it is desired to decode a part of the image, it is only necessary to decode the tile including that area without decoding the entire image, thereby improving the random access function. Can do.
[0223]
【The invention's effect】
According to the image encoding method and the image decoding method of the present invention, when decoding an image, it is possible to decode at different resolutions, or to decode only a specific tile in the image. Further, when it is desired to decode a low resolution image, only the low resolution subband information can be accessed to quickly reproduce the reduced overall image.
[0224]
Furthermore, by managing information indicating the storage position of the encoded information corresponding to each tile at a position independent of the encoded information, the position of the target tile can be easily calculated. Moreover, even if it does not decode all encoding information, the target tile can be quickly decoded by using the management information arranged independently of the encoding information.
[Brief description of the drawings]
FIG. 1 shows image coding according to the first embodiment of the present invention.MethodTheTo explainIt is a block diagram.
[Fig. 2] Image coding according to the first embodiment of the present invention.MethodIt is explanatory drawing explaining these.
[Fig. 3] Image coding according to the first embodiment of the present invention.MethodIt is explanatory drawing which shows an example of the bit stream in.
FIG. 4 is an image encoding according to the first embodiment of the present invention.MethodIt is explanatory drawing which shows another example of the bit stream in.
FIG. 5 is an image coding according to the second embodiment of the present invention.MethodIt is explanatory drawing explaining these.
[Fig. 6] Image decoding according to Embodiment 3 of the present invention.MethodTheTo explainIt is a block diagram.
[Fig. 7] Image decoding according to the fourth embodiment of the present invention.MethodIt is explanatory drawing explaining these.
[Fig. 8] Image coding according to Embodiment 5 of the present invention.MethodTheTo explainIt is a block diagram.
[Fig. 9] Image coding according to Embodiment 5 of the present invention.MethodIt is explanatory drawing explaining these.
[Fig. 10] Image coding according to Embodiment 6 of the present invention.MethodIt is explanatory drawing explaining these.
[Fig. 11] Image coding according to Embodiment 7 of the present invention.MethodTheTo explainIt is a block diagram.
FIG. 12 is an image encoding according to a seventh embodiment of the present invention.MethodIt is explanatory drawing explaining these.
FIG. 13 is an image decoding according to the eighth embodiment of the present invention.MethodTheTo explainIt is a block diagram.
FIG. 14 is an image decoding according to the ninth embodiment of the present invention.MethodTheTo explainIt is a block diagram.
FIG. 15 is an image encoding according to the tenth embodiment of the present invention.MethodTheTo explainIt is a block diagram and explanatory drawing explaining the operation | movement.
FIG. 16 is an image decoding according to the eleventh embodiment of the present invention.MethodTheTo explainIt is a block diagram and explanatory drawing explaining the operation | movement.
FIG. 17 is an image encoding according to a twelfth embodiment of the present invention.MethodExampleTo explainIt is a block diagram.
FIG. 18 is an image encoding according to a twelfth embodiment of the present invention.MethodAnother example ofTo explainIt is a block diagram.
FIG. 19 is an image encoding according to a twelfth embodiment of the present invention.MethodAnother example ofTo explainIt is a block diagram.
FIG. 20 is an image decoding according to the thirteenth embodiment of the present invention.MethodExampleTo explainIt is a block diagram.
FIG. 21 shows image decoding according to the thirteenth embodiment of the present invention.MethodAnother example ofTo explainIt is a block diagram.
FIG. 22 is an image decoding according to the thirteenth embodiment of the present invention.MethodAnother example ofTo explainIt is a block diagram.
FIG. 23: Image coding according to the fourteenth embodiment of the present inventionMethodExampleTo explainIt is a block diagram.
[Fig. 24] Image coding according to Embodiment 14 of the present invention.MethodExample of bitstream inTo explainIt is explanatory drawing.
FIG. 25: Image coding according to the fourteenth embodiment of the present inventionMethodAnother example ofTo explainIt is a block diagram.
[Fig. 26] Image coding according to Embodiment 14 of the present invention.MethodAnother example ofTo explainIt is a block diagram.
FIG. 27: Image coding according to the fourteenth embodiment of the present inventionMethodAnother example ofTo explainIt is a block diagram.
FIG. 28 shows image decoding according to the fifteenth embodiment of the present invention.MethodTheTo explainIt is a block diagram.
FIG. 29 is an image encoding according to the sixteenth embodiment of the present invention.MethodExampleTo explainIt is a block diagram and explanatory drawing explaining the operation | movement.
FIG. 30 is an image encoding according to the sixteenth embodiment of the present invention.MethodAnother example ofTo explainIt is a block diagram.
FIG. 31 is an image decoding according to the seventeenth embodiment of the present invention.MethodTheTo explainIt is a block diagram.
FIG. 32 shows conventional technology.To explainIt is a block diagram and explanatory drawing explaining the operation | movement.
FIG. 33 is a block diagram showing a conventional technique.
FIG. 34 is a block diagram showing a conventional technique.
FIG. 35 is an explanatory diagram for explaining a conventional technique.
FIG. 36 is a block diagram showing a conventional technique.
FIG. 37 is an explanatory diagram for explaining a conventional technique.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 101 ... Tile division part, 102 ... Wavelet transformation part, 103 ... Quantization part, 104 ... Entropy coding part, 105 ... Wavelet transformation coding part, 106 ... Management information generation part, 107 ... Encoded data integration part, 401 ... Management information separation unit, 402 ... encoded data extraction unit, 403 ... entropy coding unit, 404 ... inverse quantization unit, 405 ... inverse wavelet transform unit, 406 ... wavelet transform decoding unit, 407 ... tile concatenation unit, 501 ... tile Dividing unit, 502 ... Ambient pixel adding unit, 503 ... Wavelet transform unit, 504 ... Quantizing unit, 505 ... Entropy coding unit, 506 ... Wavelet transform coding unit, 507 ... Management information generating unit, 508 ... Integrated encoded data , 701... Wavelet transform unit, 702... Tile configuration unit, 703. Entropy encoding unit, 705 ... wavelet transform encoding unit, 706 ... management information generation unit, 707 ... encoded data integration unit, 901 ... management information separation unit, 902 ... encoded data extraction unit, 903 ... entropy decoding unit, 904 ... dequantization unit, 905 ... inverse wavelet transform unit, 906 ... wavelet transform decoding unit, 907 ... tile integration unit, 1001 ... management information separation unit, 1002 ... encoded data extraction unit, 1003 ... entropy decoding unit, 1004 ... inverse Quantization unit, 1005... Wavelet transform coefficient rearrangement unit, 1006... Inverse wavelet transform unit, 1007... Wavelet transform decoding unit, 1101... Wavelet transform decoding unit, 1102. 1701, 1801, 2101 2201 ... Tile wavelet encoding unit, 1602, 1706, 1802, 1905, 2005, 2105 ... Flag generation unit, 1603, 2203, 2302, 2802, 2903 ... Management information generation unit, 1604, 1703, 1803, 1903, 2004, 2103 , 2605 1st switch, 1605, 1704, 1804, 1904, 2104, 2604 ... 2nd switch, 1606, 1805 ... 3rd switch, 1607, 1702, 1807 ... wavelet encoding unit, 1608 ... first wavelet encoding unit 1705 ... 2nd wavelet encoding unit, 1806 ... 4th switch, 2204, 2503 ... encoded data combining unit, 1901, 2001, 2601 ... tile wavelet decoding unit, 1902, 2002, 2102 ... wavelet ..., 1906... 1st inverse wavelet transform unit, 2003... 2nd inverse wavelet transform unit, 2202 and 2905... ID creation unit, 2301... Data amount measurement unit, 2401. ... encoded data storage buffer, 2502 ... management information storage buffer, 2602, 3002 ... data skip control unit, 2603 ... decoding tile determination unit, 2801 ... peripheral tile ID determination unit, 2901 ... data amount storage unit, 2902 ... relative position Calculation unit, 3001 ... buffer, 2606 ... management information separation unit, 2904 ... information storage buffer, 3201, 3205, 3209, 3213 ... tile division unit, 3204, 3208, 3212 ... 1/2 reduction unit, 3202, 3206, 3210, 3214 ... JPEG compression unit, 3203, 3207, 3211, 215: Encoded data integration unit, 3301 ... Wavelet transform unit, 3302 ... Quantization unit, 3303 ... Entropy decoding unit, 3304 ... Wavelet transform coding unit, 3401, 3414, 3426 ... Horizontal low-pass filter, 3402, 3415, 3427 ... horizontal high-pass filters, 3403, 3405, 3416, 3434, 3428, 3430 ... vertical low-pass filters, 3404, 3406, 3417, 3419, 3429, 3431 ... vertical high-pass filters, 3407 to 3412, 3420 to 3425, 3432 to 3437: 1/2 sub-sampling unit, 3613: sub-band of horizontal low band / vertical low band, 3601 ... entropy decoding unit, 3602 ... inverse quantization unit, 3603 ... inverse wavelet transform unit, 3604 Wavelet transform decoding unit, 3701 ... filter application pixel, 3702 ... filter calculation target range.

Claims (3)

Dividing the image data into tiles of N pixels × M pixels and outputting N pixels × M pixels in the tile as encoding target data corresponding to each tile;
Extrapolating predetermined data around the encoding target data corresponding to each tile to perform subband division, and independently performing wavelet encoding on each tile;
Generating management information for independently decoding arbitrary tiles in units of subbands;
Adding the management information to the encoded information to generate a bitstream;
Arrangement wherein the management information includes information indicating a position in the bit stream of the coded information for each tile, seen including a management-information identifying each subband, are summarized in a position independent from said coded information picture coding method characterized in that it is. Dividing the image data into tiles, tiles and resolution of each tile type and c Eburetto coded coding information independently, a bit stream composed of a management information for managing the encoded information requires An image decoding method for decoding a decoded image according to
The management information includes the size of encoding information corresponding to each tile or resolution, and is collectively arranged at a position independent of the encoding information,
Analyzing the storage position of the encoded information corresponding to the tile or resolution to be decoded based on the size;
Performing wavelet decoding based on the storage location ;
And a step of connecting the decoded image of the wavelet decoded tile units,
Image decoding method. The image decoding method according to claim 2, wherein
  The step of analyzing the storage position of the encoding information corresponding to the tile to be decoded or the resolution based on the size is an image obtained by adding the storage position of the encoding information corresponding to the tile to be decoded or the resolution to the size Decryption method.

Images