JP2000134623A - Data compressor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To limit an amount of output code in an image compressor. SOLUTION: On the basis of an upper limit value of an output code amount of a whole image set to a register 110, a upper limit calculator 111 decides the upper limit value of the output code amount of each tile and specifies it to a tag processing part 102. Processor cores 101a to 101d execute in parallel compression processing of four components of picture data by the tile. The tag processing part 102 treats a leading part of an output code of each processor core 101 up to the upper limit value of the output code amount of each tile as effective codes, prepares and outputs a code stream with a tag.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for compressing two-dimensional data such as an image, and more particularly to a data compression technique based on a combination of a lossless wavelet transform process, a context model process, and an FSM encoding process.

[0002]

2. Description of the Related Art As a compression method for two-dimensional data such as an image, a data compression method based on a combination of a two-dimensional reversible wavelet transform process, a context (context) model process, and an FSM encoding process (referred to as a wavelet transform compression method for convenience). )It has been known. As a well-known document relating to the wavelet transform compression method, Japanese Patent Application Laid-Open No. 8-116
265 and JP-A-9-112168. The FSM encoding process is a finite state machine (FSM) based entropy encoding process, for which the encoder / decoder is called an FSM coder.

Normally, an image or the like is divided into two-dimensional areas called tiles, and compression is performed for each tile. The compressed data is output as a code stream of a predetermined format.

[0004]

For example, when an image captured by a digital camera is compressed and stored in a memory card or the like, the code amount (compressed data amount) per image is often limited. Such a limit on the amount of code
Not limited to the example described above, it is often necessary when storing and transmitting data.

An object of the present invention is to provide a data compression apparatus of a wavelet transform compression system suitable for an application requiring such a restriction on the code amount.

Here, in order to facilitate understanding of the objects and effects of the present invention, several methods for limiting the code amount in the wavelet transform compression method will be examined.

One conceivable method is to encode the data after the wavelet transform processing in order from the data having the highest importance, and if the data having the predetermined importance is encoded, the remaining data is discarded. Will.
In this method, since the actual amount of code is not known until the encoding is completed, if the amount of code exceeds the limit at the end of encoding, up to which importance data is to be encoded It is necessary to re-specify and perform the encoding again, which increases the processing time.

Another conceivable method is to sequentially encode data having high importance, sequentially measure the amount of generated codes, and when a predetermined amount of codes is reached, code generated thereafter is replaced with a predetermined code amount. It would be a way to throw it away. This method could be used if the image is not tiled. However, when processing an image of a certain size, it is common to divide the image into a plurality of tiles and encode them sequentially in tile units. When performing tile division, if the code amount is measured during encoding and the code after reaching the predetermined code amount is discarded, the data of tiles after a certain tile will be completely lost. ,
You cannot restore a satisfactory image from the codestream.

One method of avoiding such inconvenience is to encode all tiles and store the generated codes in a memory. At the stage when the encoding is completed, the entire code amount is grasped, and then the code of each tile is quantized so that the entire code amount does not exceed the limit. However, it is necessary to prepare a large memory for storing codes for one image. One of the purposes of dividing an image into tiles is to suppress an increase in the amount of memory required for processing, and it is not preferable to require such a large memory. Further, there is a problem that it takes a considerable time to perform a process of discarding a part of data of each tile with respect to a code stream which is sequential data of each tile.

According to the present invention, as will be apparent from the following description, it is possible to appropriately limit the amount of codes while avoiding the disadvantages such as the above-described methods.

[0011]

According to a first aspect of the present invention, there is provided a wavelet transform compression type data compression apparatus for compressing two-dimensional data such as an image and outputting a code stream. First means for holding an upper limit value of the output code amount for the entire data;
Means for determining the upper limit value of the output code amount of each tile based on the upper limit value of the output code amount held in the first means, performing compression processing on a tile-by-tile basis, and determining the output code amount of each tile. That is, the upper limit is determined by the second means.

According to a second aspect of the present invention, there is provided a data compression apparatus of a wavelet transform compression system for compressing two-dimensional data such as an image and outputting a code stream. First means for holding the upper limit of the amount, second means for holding the weight of each component of the two-dimensional data set from outside, and the upper limit of the output code amount held by the first means And a third means for determining an upper limit value of an output code amount for each component of each tile based on the weight of each component held in the second means, performing compression processing in tile units, The purpose is to limit the output code amount of each component of each tile to the upper limit determined by the third means.

According to a third aspect of the present invention, there is provided a data compression apparatus of a wavelet transform compression system for compressing two-dimensional data such as an image and outputting a code stream. First means for holding the upper limit of the amount, second means for holding the weight of each tile set from the outside, the upper limit of the output code amount held by the first means, and the second A third means for determining an upper limit value of the output code amount of each tile based on the weight of each tile held in the means, performing compression processing on a tile-by-tile basis, and In other words, the upper limit is determined by the third means.

According to a fourth aspect of the present invention, there is provided a data compression apparatus of a wavelet transform compression system for compressing two-dimensional data such as an image and outputting a code stream. First means for holding the upper limit of the amount, second means for holding the weight of each component of the two-dimensional image data set from the outside, and second means for holding the weight of each tile set from the outside 3 means, the upper limit of the output code amount held by the first means, the weight of each component held by the second means, and the weight of each tile held by the third means And a fourth means for determining an upper limit value of an output code amount of each component of each tile, performing compression processing on a tile-by-tile basis, and outputting an output code amount of each component of each tile. It is to limit the upper limit value determined by means of the fourth.

The feature of the invention of claim 5 is that the features of claims 1, 2, and
5. The data compression device according to 3 or 4, wherein a wavelet transform unit for a reversible wavelet transform process, a context model unit for a context model process,
Compression processing of a plurality of components of two-dimensional data is performed in tile units by a plurality of processor cores each composed of an FSM coder for FSM encoding processing.

The feature of the invention of claim 6 is that the features of claims 1 to 5
In the data compression device according to any one of the above, when the output code amount of each tile reaches an upper limit value, the compression process for the tile is stopped, the output operation of the code stream of the tile and the data of the next tile are performed. This is to start the input operation at the same time.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image data compression apparatus according to an embodiment of the present invention will be described below with reference to the accompanying drawings. For simplification of description, the same reference numerals are used for the same or corresponding parts in a plurality of drawings.

FIG. 1 is a block diagram showing an example of an image data compression device according to the present invention. The image data compression apparatus 100 divides a color image into tiles, performs a compression process by a wavelet transform compression method on a tile basis, and outputs a code stream. To speed up the process, the four components A and B of the image are used. , C and D are executed in parallel.

For this parallel processing, the image data compression apparatus 100 includes four processor cores 101a, 101a, one for each component of a color image.
01b, 101c, and 101d operate in parallel, and code output from each processor core 101 is transmitted to the tag processing unit 10b.
This is a basic configuration in which the data is unified into two and output as one code stream to which a tag is added.

As shown in FIG. 2, the code stream includes a main header (Main header), a tile header (tile header) for each tile, and tile data (tile data) as a compression code of an image in order. It is a side-by-side structure. As shown in FIG. 4, the content of each tile data is obtained by arranging compression codes of components A, B, C, and D in that order. Before the main header,
At the end of the code stream, markers are placed between each header and between the header and the tile data, but these are omitted in the figure.

The components of the color image processed here include, for example, R (red), G (green), B (blue), and alpha components of the primary color system (information such as the depth of three-dimensional CG). Component for preservation). Alternatively, a complementary color Y (yellow) component, an M (magenta) component, a C (cyan) component, and a K (black) component are used. However, it is not limited to this.

Each of the processor cores 101 has the same configuration. For each tile, a wavelet transform unit 103 for performing a two-dimensional reversible wavelet transform process, a context model unit 104 for performing a context model process, and an FS
And an FSM coder that performs M encoding. In each processor core 101, the wavelet transform unit 103 decomposes the input component data of each tile into frequency band signals by two-dimensional reversible wavelet transform. The context model unit 104 determines target bits for the frequency band signal in the order according to the specified alignment information, and generates a context from the arrangement of peripheral bits of the target bits.
The FSM coder 105 generates a code from the context and the target bits by probability estimation.

As described above, the tag processing unit 102 unifies the codes output from the respective processor cores 101 to create a code stream to which a tag is added. At this time, the output code amount is determined for each tile. Restrict. In connection with the limitation of the output code amount, the data compression device includes an upper limit setting register 110 and an upper limit calculator 111. In the upper limit setting register 110, the upper limit value of the output code amount of one entire image is externally set. Upper limit calculator 111
Calculates the upper limit value of the output code amount of each tile based on the upper limit value set in the upper limit setting register 110, and designates the upper limit value of the output code amount to the tag processing unit 102.

For example, the output code amount upper limit set in the upper limit setting register 110 is 160 KByte, and the number of tiles constituting one image is set to 16 (the number of tiles is set in the upper limit calculator 111 as a default value). And when
There is a case where the upper limit calculator 111 is specified from outside via a register or the like. The latter has a variable number of tiles). In this case, the upper limit calculator 111 calculates the upper limit of the code amount of each tile as 160 KByte / 16 = 10 KByte (note that since the header of the tile is short, the calculation is ignored, and the same applies hereinafter). In the tag processing unit 102, of the codes of each tile output from each processor core 101, codes up to the first 10 KBytes are treated as valid, and a code stream as shown in FIG. 2 is created and output. Note that the amount of code assigned to each component is equal.

As described above, the image data compression apparatus 10
0 is a code in which the compression processing of each component is executed in parallel in tile units by the lossless wavelet transform processing, the context model processing and the FSM encoding processing, and the output code amount of each tile is limited (equivalent to quantization). This is a configuration that creates and outputs a stream. Therefore,
There is no need for a large memory for temporarily storing the code data of the entire image. Further, even when the code amount after compression exceeds the upper limit, it is not necessary to perform the compression process again.
Further, since the output code amount is limited for each tile, data of a specific tile is not completely lost. Further, the processing time from input of image data to output of a code stream can be reduced as compared with a method of temporarily storing code data of the entire image and then discarding a part of each tile data.

The image data compression apparatus 100 is capable of parallel processing of a maximum of four components.
Processing of up to three components is also possible. For example,
In the case of a monochrome image represented by only one component, the processing can be executed by only one processor core 100. It goes without saying that an image data compression apparatus having a configuration other than such parallel processing is also included in the present invention.

FIG. 3 is a block diagram showing another example of the image data compression device according to the present invention. The basic configuration and operation of the image data compression apparatus 200 are the same as those of the image data compression apparatus 100. The difference is that a component weight register 112 is provided in addition to the upper limit setting register 110 in relation to the code amount restriction. That is, the operation of the upper limit calculator 111 and the operation of the tag processing unit 102 are partially different with respect to the code amount restriction. Hereinafter, the difference will be mainly described.

In the upper limit setting register 110, the upper limit of the output code amount of the entire image is externally set as described above. The component weight register 112 stores the relative weight (relative importance) of each of the components A, B, C, and D.
Is set externally. The upper limit calculator 111 determines an upper limit value of the output code amount of each tile and each component of each tile based on the information set in each of the registers 110 and 112 and the number of tiles (a default value or a value set from outside). Is calculated, and the upper limit value of the output code amount is calculated and specified to the tag processing unit 102.

For example, the upper limit of the code amount of the entire image is 1
60KByte, 16 tiles, components A, B,
It is assumed that the weights of C and D are 4: 3: 2: 1. The upper limit calculator 111 sets the upper limit value of the output code amount of each tile to 160 KBy.
Calculate te / 16 = 10KByte. Next, the upper limit value of the output code amount of each tile is apportioned according to the weight of each component, and the upper limit value of the code amount of each component is calculated.
In this example, the upper limit of the output code amount of each component of A, B, C, and D is A = 4 KByte, B = 3 KByte, and C = 2.
KByte, D = 1KByte is calculated. In this case, for each tile, the tag processing unit 102 treats, from the beginning of the code of each component output from each processor core 101, up to the upper limit value as valid data, and creates a code stream. Output. FIG. 4 shows the relationship between the configuration of the code stream and the upper limit of the output code amount of each tile and each component.

There are cases where it is desired to change the reproduction accuracy for each component depending on the type of image or the user's desire. According to the image data processing device 200, the output code amount can be appropriately limited in a situation where such component weighting is required. Further, it is possible to avoid an increase in memory and processing time and a loss of specific tile data, similarly to the image data processing apparatus 100.

The image data compression device 200 can also process a single component monochrome image. In this case, since the code amount upper limit value of the component and the code amount upper limit value of the tile data are the same value,
The result of the code amount limitation is substantially the same as that of the image data compression processing device 100.

FIG. 5 is a block diagram showing another example of the image data compression device according to the present invention. The basic configuration and operation of the image data compression apparatus 300 are the same as those of the image data compression apparatus 100. The difference is that a tile weight register 114 is provided in addition to the upper limit setting register 110 in relation to the code amount restriction. That the upper limit calculator 111
And the operation of the tag processing unit 102 is partially different with respect to the output code amount limitation. Hereinafter, the difference will be mainly described.

In the upper limit setting register 110, the upper limit of the output code amount of the entire image is externally set as described above. In the tile weight register 114, the relative weight (relative importance) of each tile is set from the outside. The upper limit calculator 111 calculates the upper limit value of the output code amount of each tile based on the information set in the registers 110 and 114 and the number of tiles (default value or a value set from the outside), and tags it with a tag. It is specified to the processing unit 102.

For example, the upper limit of the code amount of the entire image is 2
70KByte, 9 tiles, weight of each tile 4: 3:
It is assumed that 2: 4: 3: 2: 4: 3: 2. The upper limit calculator 111 apportions 270 KByte according to the weight of each tile, and sets the upper limit of the output code amount of each tile to 40 KByte, 3
0KByte, 20KByte, 40KByte, 30KByte, 20KByt
Calculate as e, 40KByte, 30KByte, 20KByte. In this case, the tag processing unit 102 treats the code of each tile from the beginning to the upper limit value specified by the upper limit calculator 111 as a valid code, and converts a code stream including only valid codes into a code stream composed of only valid codes. Create and output. The output code amount assigned to each component in each tile is equal. The structure of the codestream,
FIG. 6 shows the relationship between the output code amount upper limit values of each tile.

Depending on the type or use of the image, there is a case where it is desired to increase the reproduction accuracy of a specific tile in the image and lower the reproduction accuracy of other tiles. According to the image data processing device 300, it is possible to appropriately limit the output code amount in a situation where such tile weighting is required. Further, it is possible to avoid an increase in memory and processing time and a loss of specific tile data, similarly to the image data processing apparatus 100. The image data compression apparatus 300 can also process a single component monochrome image.

FIG. 7 is a block diagram showing another example of the image data compression apparatus according to the present invention. The image data compression device 400 can limit the amount of output code in which one or both of a tile and a component are weighted or not, and can perform high-speed processing.

The image data compression device 400 includes an upper limit setting register 110, a component weight register 112, and a tile weight register 114, all of which are similar to those of the image data processing devices, in relation to the output code amount limitation.
The upper limit calculator 111 determines the upper limit value of the output code amount of the entire image according to the information of the code amount upper limit register 110,
Within this entire code amount, the tile weight register 1
14 and the component weight register 112, the upper limit value of the output code amount of each tile and each component is determined. The image data compression device 400 further includes a control device 118 that controls the operation state of each processor core 101 according to the upper limit value calculated by the upper limit calculator 111.

The upper limit value setting register 110 sets the upper limit value of the output code amount of the entire image. The weight of each component is set in the component weight register 112, but if it is not desired to weight the components, that is, if all the components are to be treated with the same importance, an equal weight is applied to all the components. Is set. The weight of each tile is set in the tile weight register 114. However, if it is not desired to weight the tiles, that is, if all the tiles are to be treated with the same importance, an equal weight is applied to all the tiles. Is set.

When the weights of the tiles are equal, the upper limit calculator 111 divides the upper limit value of the code amount set in the upper limit setting register 110 by the number of tiles (default value or a value specified from the outside) to obtain each tile. The upper limit of the output code amount for each is determined. When different weights are assigned to each tile, the output code amount of each tile is determined by apportioning the upper limit of the total output code amount according to the weight of each tile. Then, the upper limit of the output code amount of each component in each tile is determined by apportioning the upper limit of the output code amount of each tile determined in this way according to the weight of each component. After determining the upper limit value of the output code amount of each component by apportioning the upper limit value of the entire output code amount according to the component weight, the output code amount upper limit value of each component is apportioned according to the weight of each tile. Thus, the upper limit value of the output code amount of each component in each tile may be determined. The output code amount upper limit determined in this way is the
Is specified in the tag processing unit 102.

FIG. 8 is an operation timing chart of the image data compression apparatus 400. In FIG. 8, the “processing status” indicates the operation state of the entire apparatus, and its “I / O”
Input of image data in units of tiles and output of code streams are performed in parallel during the period, and compression processing in units of tiles is performed during the “encoding process” period.

In FIG. 8, “core A status” to “core D status” represent the operation states of the processor cores 101 corresponding to components A to D, and “wai”.
"t" is an input / output waiting state, "WL" is a wavelet transform processing execution state, and "CM &FSM" is a context model processing and FSM encoding processing execution state. The “input image data” indicates the state of the image data input, and the “invalid”
Indicates that valid image data is not being input, and “tile n” indicates that valid image data of tile n is being input. "Code stream" indicates the status of the code stream output, "Invalid" indicates that a valid code stream is not being output, and "Tile n" indicates that a valid code stream of tile n is being output. Represents

The operation flow will be described with reference to FIG.
The image data compression device 400 takes in the image data of the first tile 1 during the first I / O period. The capture ends at time a, and each processor core 101 immediately starts the wavelet transform processing of each component, and after that, starts the context model processing and the FSM encoding processing. The control device 118 controls each processor core 101
The amount of code output from the FSM coder 105 is measured and compared with the upper limit of the output code amount of the corresponding component of the tile 1 specified by the upper limit calculator 111. When the output code amount of a certain processor core 101 reaches the upper limit, the context model processing and the FSM encoding processing of the certain processor core 101 are stopped.
In the timing example shown in FIG. 8, the processing of the processor core 101b of the component B is stopped at the time b,
Next, at time c, the processor core 101 of the component C
c is stopped, and at time d, the components A,
The processing of the processor cores 101a and 101b of D is stopped, and the encoding processing of tile 1 is completed. That is,
The output code amount of tile 1 has reached the upper limit. Immediately, the next I / O operation is started, and the acquisition of the image data of the next tile 2 is started. In parallel with this, the tag processing unit 102 outputs the code data (code) of the tile 1 output from each processor core 101. Is output as one code stream. Thereafter, the same operation is repeated until the last tile.

As described above, the image data compression apparatus 400 can limit the amount of code according to the importance of tiles and the importance of components. Further, when the output code amount of a certain component reaches the upper limit value, ie, when the output code amount upper limit value of each tile is the maximum,
When the output code amount of each tile reaches the upper limit value, the I / O operation is started immediately, and the input of the image data of the tile and the output of the code stream are simultaneously performed. Therefore, compared to a configuration in which the data of all the components of each tile are encoded to the end and then the code stream is output, or a configuration in which the output of the code stream and the input of the image data are not performed simultaneously, the processing time per tile is greatly reduced. Can be shortened. Further, the image data compression devices 100, 20
As with 0,300, no large memory is required and data loss for a particular tile is also avoided.

Although not shown, a control device similar to the control device 118 in the image data compression device 400 in FIG. 7 is provided in the image data compression device in FIG. 3 or FIG. When the output code amount of all components of each tile reaches the upper limit, the output of the output code stream and the capture of the image data of the next tile are started at the same time, and the processing time of each tile is reduced. Further shortening can be achieved, and the image data compression apparatus having such a configuration is naturally included in the present invention.

[0045]

According to the first to sixth aspects of the present invention,
The image or the like can be compressed to a desired code amount according to the properties of the image or the like to be compressed, the storage capacity of the medium for storing the compression result, and the reproduction accuracy desired by the user. A large memory for temporarily storing the code data of the entire two-dimensional data such as an image is not required. Even when the amount of code after compression exceeds the upper limit, there is no need to perform compression processing again. Since the output code amount is limited for each tile,
There is no complete loss of data for a particular tile. Further, the processing time from input of two-dimensional data to output of a code stream can be reduced as compared with a method of temporarily storing code data of entire two-dimensional data such as an image and then discarding a part of each tile data. .

Depending on the type of image or the like, the reproduction accuracy of a specific tile or component may be increased, and the reproduction accuracy of other tiles or components may be reduced. According to the invention of claim 2, 3 or 4,
It is possible to appropriately limit the output code amount in a situation where such tiles and components need to be weighted.

According to the fifth aspect of the present invention, two-dimensional data composed of a plurality of components such as a color image can be compressed at a high speed, and the output code amount can be appropriately limited.

According to the sixth aspect of the present invention, the output of the code stream is started after the data of each tile is compressed to the end, and the input of the data of each tile and the output of the code stream are temporally separated. As compared with a configuration in which the processing is performed by a single operation, the processing time per tile can be significantly reduced, and the amount of output codes can be appropriately limited.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an example of an image data compression device according to the present invention.

FIG. 2 is an explanatory diagram of a code stream configuration and code amount limitation.

FIG. 3 is a block diagram showing another example of the image data compression device according to the present invention.

FIG. 4 is an explanatory diagram of a code stream configuration and a code amount limit.

FIG. 5 is a block diagram showing another example of the image data compression device according to the present invention.

FIG. 6 is an explanatory diagram of a code stream configuration and a code amount restriction.

FIG. 7 is a block diagram showing another example of the image data compression device according to the present invention.

FIG. 8 is an operation timing chart.

[Explanation of symbols]

101a, 101b, 101c, 101d Processor core 102 Tag processing unit 103 Wavelet transform unit 104 Context model unit 105 FSM coder 110 Upper limit setting register 111 Upper limit calculator 112 Component weight register 114 Tile weight register 118 Control device

Claims (6)

[Claims] 1. A data compression device that compresses two-dimensional data such as an image by a compression process based on a combination of a reversible wavelet transform process, a context model process, and an FSM encoding process, and outputs a code stream. First means for holding the upper limit of the output code amount for the entire two-dimensional data, and means for determining the upper limit of the output code amount of each tile based on the upper limit of the output code amount held by the first means A data compression apparatus that performs compression processing on a tile-by-tile basis and limits the output code amount of each tile to an upper limit determined by the second means. 2. A data compression device that compresses two-dimensional data such as an image by a compression process based on a combination of a reversible wavelet transform process, a context model process, and an FSM encoding process, and outputs a code stream. First means for holding the upper limit of the output code amount for the entire two-dimensional data, second means for holding the weight of each component of the two-dimensional data set from the outside, and holding in the first means A third means for determining an upper limit value of an output code amount of each component of each tile based on the output code amount upper limit value obtained and the weight of each component held in the second means, Performs compression processing, and limits the output code amount of each component of each tile to the upper limit determined by the third means. Data compression device. 3. A data compression device which compresses two-dimensional data such as an image by a compression process based on a combination of a reversible wavelet transform process, a context model process and an FSM encoding process, and outputs a code stream, which is set from outside. First means for holding the upper limit of the output code amount for the entire two-dimensional data, second means for holding the weight of each tile set from the outside, and output code amount held by the first means A third means for determining an upper limit value of the output code amount of each tile based on the upper limit value and the weight of each tile held in the second means, performing compression processing in tile units, and Characterized in that the output code amount is limited to an upper limit determined by the third means. 4. A data compression device that compresses two-dimensional data such as an image by a compression process based on a combination of a reversible wavelet transform process, a context model process, and an FSM encoding process, and outputs a code stream. First means for holding the upper limit of the output code amount for the entire two-dimensional data, second means for holding the weight of each component of the two-dimensional data set externally, and each tile set externally And the upper limit of the output code amount held by the first means, the weight of each component held by the second means, and the weight held by the third means. A fourth means for determining an upper limit value of an output code amount of each component of each tile based on a weight of each tile; A data compression device for limiting an output code amount of each component of each tile to an upper limit determined by the fourth means. 5. A plurality of processor cores each comprising a wavelet transform unit for a reversible wavelet transform process, a context model unit for a context model process, and an FSM coder for an FSM encoding process, 5. The data compression apparatus according to claim 1, wherein compression processing of a plurality of components of the two-dimensional data is executed in parallel in tile units. 6. When the output code amount of each tile reaches the upper limit value, the compression processing for the tile is stopped, and the output operation of the code stream of the tile and the input operation of the data of the next tile are started simultaneously. The data compression processing device according to claim 1, wherein: