JPH06178117A - Image encoder - Google Patents

Abstract

PURPOSE:To always obtain a constant compression ratio at every block by encoding the data of the hierarchy of gradation and the hierarchy of resolution and stopping an encoding processing at a point of time when the total amount of encoded data reaches preliminarily set code amount. CONSTITUTION:In a significant rank determination device 3, the ranking of each hierarchy when the picture element within a block is divided into the hierarchy of resolution and the hierarchy of gradation is performed based on an analysis result 8. In this case, the hierarchy of prescribed resolution and the hierachy of gradation are defined as a first rank, and the candidate of the hierarchy of the resolution and the candidate of the hierarchy of gradation where an input block 7 is available to be decoded in a state less than a prescribed encoded error, are determined respectively on the feature amount of resolution and the feature amount of gradation. By starting from the first rank hierarchy of resolution and the hierarchy of gradation, the ranking is performed for the candidate of the hierarchy of resolution and the candidate of the hierarchy of gradation. At this point, a rank 9 and the output 10 of a gradation/ resolution computerization device 4 are encoded by an information source encoder 5, encoded data amount is added to it, and it is terminated by reaching an aimed code amount.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for coding image information.

[0002]

2. Description of the Related Art For example, in a printer, raster image data to be output is temporarily stored in a page memory. Further, in the image editing apparatus, the image data to be edited or the edited image data is also temporarily stored in the page memory. The required capacity of the page memory increases as the image size increases, the resolution increases, and the number of gradations per pixel increases. For example, when storing a full-color image of A3 size of 167.7 million colors / pixel at a resolution of 16 pixels / mm, the memory capacity is 96 Mbytes. When the memory capacity increases in this way, the cost of the page memory increases, and it takes time to read and / or write the image data, which causes a problem of a long processing time.

As a solution to such a problem, it has been considered to reduce the page memory capacity by highly efficiently encoding the image data in a form that allows the encoded data to be edited. The following three points are required for such encoding.

A constant compression rate is possible for each predetermined image unit.

Local encoding / decoding is possible for each predetermined image unit.

Uniform encoding / regardless of image type
Decryption is possible.

First, since the page memory has a finite capacity, it is necessary to be able to encode each predetermined image unit at a preset compression rate without depending on image data (request). In this case, the predetermined image unit needs to be at least a page unit. Further, in order to edit the following code data as it is, the predetermined image unit needs to be a unit obtained by dividing a page into several image areas. For example, it is a block unit of 8 pixels × 8 lines.

Secondly, in order to edit the coded data as it is, it is necessary that each predetermined image unit can be independently coded / decoded (request).

Thirdly, since encoding / decoding is performed on the page memory, it is desirable that the processing can be performed at high speed and at a constant speed (request).

In the conventional image coding apparatus for storage / transmission, visual redundancy of image data as much as possible, and
Since it is necessary to suppress the statistical redundancy, the compression rate varies due to the variation of the redundancy for each image data. In addition, there is a tendency to introduce more advanced encoding processing, and it is difficult to perform encoding / decoding independently for each predetermined image division unit.
Furthermore, it has been difficult to satisfy the requirements (1) to (3) because the introduction of the adaptive processing causes the amount of calculation required for the encoding / decoding processing to fluctuate significantly according to the fluctuation of the redundancy of each image data.

As a technique satisfying the requirements (1), there is a technique filed by the present applicant as Japanese Patent Application No. 3-204436. Hereinafter, the configuration of the image encoding device proposed in the specification of the same application will be described with reference to FIG. The device is
Blocking means 201 that samples the image data and divides it into blocks of m × n pixels (m and n are positive integers) composed of a plurality of pixels, average value calculating means 202 that calculates an average value in the input block, and input block An average value separating means 203 for subtracting the average value from each pixel in the pixel, a means 204 for analyzing the resolution and gradation feature amount of the block after the average value separation,
A plurality of pixel thinning shapes and pixel thinning rates set in advance based on the analysis result, and mode determining means 205 for independently determining the number of gradations of pixels in the average value separation block, and the determined pixel thinning shapes and pixels Resolution approximation means 206 for thinning out pixels in the average value separation block according to the thinning rate
And a gradation approximation unit 207 that quantizes the thinned pixels with the gradation number determined by the mode determination unit, and an average value, a mode determination result, and the output of the gradation approximation unit are multiplexed to form code data. And multiplexing means 208 for

Next, the operation of the image coding apparatus shown in FIG. 12 will be described. The image is sampled and divided into m × n pixel input blocks composed of a plurality of pixels by the blocking unit 201. The average value in the divided input block is obtained by the average value calculating means 202, and the average value separating means 203 subtracts the average value from the pixels in the input block to obtain the average value separating block. Next, the feature amount analysis unit 204 analyzes the resolution and tone feature amounts of this average value separation block. Based on the analysis result, a plurality of pixel thinning shapes and pixel thinning rates set in advance, and the number of gradations of pixels in the average value separation block are determined by the mode determining means 205. Pixels in the average value separation block are thinned by the resolution approximating means 206 according to the pixel thinning shape and the pixel thinning rate determined by the mode determining means 205, and then the mode determining means 205.
The tone approximation unit 207 quantizes the number of tones determined in. The average value, the mode determination result, and the output of the gradation approximating means are multiplexed by the multiplexing means 208 to form code data.

In the conventional image coding apparatus shown in FIG. 12, a combination having a predetermined compression rate is selected from among combinations of pixel thinning-out shapes, pixel thinning-out rate candidates, and gradation number candidates. By determining, the code amount is controlled to be constant in input block units.

The data which has been subjected to the pixel thinning and quantization described above still has a statistical redundancy, but the information source coding for suppressing the redundancy is not applied. Therefore, it is expected that the coding efficiency can be improved by adding the information source coding means to this device. However, when the information source coding is added, the code amount fluctuates due to the local fluctuation of the redundancy, so that the code amount control becomes necessary.

As a conventional technique using this code amount control, there is, for example, the encoding system described in D-45, 1989, SINK. The configuration will be described below with reference to FIG. In this method, a block forming unit 301 that samples an image and divides it into blocks of 8 × 8 pixels composed of a plurality of pixels, and a discrete cosine transforming unit (discrete DCT transforming unit in the figure) that performs discrete cosine transform of pixel values in the block. ) 302, and a conversion coefficient storage unit 303 for temporarily storing the converted conversion coefficient
, Quantizing means 304 for quantizing the transform coefficient in a predetermined quantizing step, variable length coding means 305 for variable length coding the output of the quantizing means 304, and measuring the data amount of the coded code. Then, the measurement / estimation means 306 for estimating the quantization step having a preset code amount based on the measurement result.

Next, the operation will be described with reference to FIG. The image data is sampled and divided into m × m pixel input blocks including a plurality of pixels by the blocking unit 301. The pixel values in the divided input block are subjected to discrete cosine transform by the discrete cosine transform means 302. The converted transform coefficient is temporarily stored in the transform coefficient storage unit 303. The transform coefficient stored in the transform coefficient storage means 303 is quantized by the quantizing means 304 by a predetermined quantizing step, and then is variable length coded by the variable length coding means 305. At this time, the data amount of the coded code is measured by the measurement / estimation means 306, and the quantization step that provides the preset code amount is estimated based on the measurement result. The transform coefficient stored in the storage means is quantized again in this inferred quantization step, and then variable length encoded in the variable length encoding means 305. Hereinafter, the operations of quantization, variable length coding, and measurement / estimation are repeated until the code amount set in advance is reached.

This method is based on the JPEG method, and the relationship between the quantization step of the transform coefficient, which is a parameter that influences the code amount, and the code amount is investigated in advance, and the code amount is controlled using this. Is. In this method, a control error of the code amount always occurs, so that there is a problem that the following measures need to be taken.

It is necessary to provide the target code amount with a sufficient margin.

It is necessary to change the parameters and repeat the encoding until the error becomes sufficiently small.

It is necessary to consider what to do when the code amount overflows.

The problem is contrary to the basic requirement at the time of encoding for improving the encoding efficiency, and the problem is contrary to the requirement for performing uniform encoding / decoding processing regardless of the type of image.

[0022]

The conventional technique shown in FIG. 13 is required as an encoding device applied to a page memory.
However, because the code data after resolution approximation and gradation approximation has statistical redundancy, the coding efficiency is higher than that of the conventional high-efficiency coding method for storage and transmission. There was a problem of low. If information source coding is added to the conventional technique shown in FIG.

An object of the present invention is to control a constant code amount for each input block and to improve coding efficiency.

[0024]

In order to achieve the above-mentioned object, an image coding apparatus of the present invention samples an image into an input block of m × n pixels (m and n are positive integers) and is composed of a plurality of pixels. Blocking means for dividing, analyzing means for analyzing the resolution and gradation feature amount of pixels in the input block, and each layer when the pixels in the input block are divided into resolution hierarchy and gradation hierarchy Of the input block based on the analysis result of the analysis unit, and the pixels in the input block are divided into the hierarchy of the resolution and the hierarchy of the gradation according to the order determined by the importance order determining unit. And a gradation / resolution information layering means for outputting the same, and an encoding means for encoding the rank and the output of the gradation / resolution information layering means and outputting code data.

[0025]

The image is sampled and divided by the blocking means into m × n pixel input blocks consisting of a plurality of pixels. Next, the gradation / resolution analysis means analyzes the resolution and gradation characteristic amount of the block. When the pixels in the block are divided into a hierarchy of resolution and a hierarchy of gradations, the important rank determining means ranks each hierarchy based on the analysis result. At the time of this ranking, for example, the hierarchy of a predetermined resolution and the hierarchy of the gradation are set as the first rank, and the input block is a candidate of the hierarchy of the resolution that can be decoded with a preset coding error or less and the gradation. Candidates for each layer are independently obtained based on the characteristic amount of resolution and the characteristic amount of gradation, starting from the layer of the first order of resolution and the layer of gradation, and the candidate of the layer of resolution and the layer of gradation. Ranking is performed toward the candidates. In the gradation / resolution information layering means, the pixels in the block are divided into a resolution layer and a gradation layer and output according to the order. The output of the rank / gradation / resolution information layering means is encoded by the information source encoding means. In the information source coding means, for example, the amount of code data is added at the same time as the encoding process, and the encoding process is terminated when the amount reaches a preset target code amount, and the amount of code data is equal to or less than the target code amount. Code data of the data amount is output.

At this time, both the visual redundancy and the statistical redundancy of the input block are suppressed, and it is not necessary to provide the target code amount with a margin in consideration of the control error. High coding is possible. at the same time,
Each m × n pixel block is encoded with a fixed code amount, and each m × n pixel block can be independently encoded / decoded. Further, since the iterative process for controlling the code amount and the exceptional process when the code amount exceeds the target value are unnecessary, the uniform encoding / decoding process is performed. Furthermore, the pixels in the block are divided into a hierarchy of resolution and a hierarchy of gradations, and encoding is performed in order from the most important hierarchies in order to reduce encoding errors. It is possible to minimize the deterioration.

[0027]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The features of the present invention will be specifically described below based on embodiments with reference to the drawings.

First, the basic structure and operation of the present invention will be described.

As shown in FIG. 1, according to the present invention, the image data 6 is sampled and m × n pixels (m, n) composed of a plurality of pixels are sampled.
Is a positive integer) Blocking means 1 for dividing into blocks 7
And a gradation / resolution analysis means 2 for analyzing the resolution and gradation feature amount of the pixels in the block, and the ordering of each hierarchy when the pixels in the block are divided into a resolution hierarchy and a gradation hierarchy. Based on the analysis result 8, the important order determining unit 3 and the gradation / output which divides the pixels in the block into the resolution hierarchy and the gradation hierarchy according to the determined resolution hierarchy and the gradation hierarchy 9 The resolution information layering means 4, the output 9 of the important order determining means 3 and the output 10 of the gradation / resolution information layering means 4 are source coded to output code data 11, and the amount of code data is added. It has an information source coding means 5 for ending the coding process when the preset code amount is reached.

In the coding apparatus shown in FIG. 1, the image data 6 is sampled and divided by the blocking means 1 into an input block 7 of m × n pixels composed of a plurality of pixels. Next, the gradation / resolution analysis means 2 analyzes the resolution and gradation feature amount of the block 7. When the pixels in the block are divided into a resolution hierarchy and a gradation hierarchy, the importance ranking determination means 3 ranks each hierarchy based on the analysis result 8. In this case, the hierarchy of the predetermined resolution and the hierarchy of the gradation are set as the first rank, and the candidate of the hierarchy of the resolution and the candidate of the gradation of the resolution which can be decoded by the input block 7 with a predetermined coding error or less are set as the resolution. Independently, based on the feature amount and the gradation feature amount, starting from the resolution hierarchy and the gradation hierarchy of the first rank, and proceeding toward the resolution hierarchy candidate and the gradation hierarchy candidate. Attached. In the gradation / resolution information layering means 4, the pixels in the block are divided into the resolution layer and the gradation layer in accordance with the order 9 and output. The rank 9 and the output 10 of the gradation / resolution information layering means 4 are encoded by the information source encoding means 5. In the information source coding means 5, the amount of code data is added at the same time as the coding process, and when the amount of code data reaches a preset target code amount, the coding process is discontinued and data having a target code amount or less is added. The quantity code data 11 is output.

The configuration of each part of the above-mentioned encoding device will be described below in detail with reference to FIGS. 2 to 7.

The gradation / resolution analysis means 2 shown in FIG. 2 is a gradation analyzer 14 for analyzing the gradation feature quantity of the input block 7.
, A resolution analyzer 15 for analyzing the feature quantity of the resolution of the input block 7, a gradation analysis result 17 output by the gradation analyzer 14 and a resolution analysis result 18 output by the resolution analyzer 15 are multiplexed to obtain an analysis result. It is composed of a multiplexer 16 which outputs 8.

As shown in FIG. 3, the gradation analyzer 14 is composed of a dispersion calculator 19 which calculates and outputs a dispersion value of pixels in the input block 7.

The resolution analyzer 15 is, as shown in FIG.
A block divider 24 that divides the input block 7 by a positive integer ratio j and an average value of the pixels in the divided block 25, and an average value separation that outputs by subtracting this average value from each pixel in the block 25 Unit 26, a first vector set 28 that stores a representative shape block obtained in advance, an inner product of the average value separation block 36 and the representative shape block 37 stored in the first vector set 28, and the positive and negative of the inner product is calculated. A first inner product calculator 27 that outputs a code, a first index holder 29 that holds the positive and negative signs 38 of the inner product as an index, a second vector set 31 that stores a set of previously determined representative shape blocks, Representative shape block 4 indicated by index 39 among the average shape separation block 36 and the representative shape block stored in the second vector set 31. A second inner product calculator 30 that calculates the inner product of and the positive and negative signs 41 of the inner product, and a second index holder 32 that holds the positive and negative signs 41 of the inner product and the index 39 as a new index, and An inner product of a third vector set 34 that stores a set of representative shape blocks, an average value separation block 36, and a representative shape block 43 indicated by an index 42 in the representative shape blocks stored in the third vector set 34. From the third inner product calculator 33 that calculates the positive and negative signs 44 of the inner product and the third index holder 35 that holds the positive and negative signs 44 and the index 42 of the inner product as new indexes and outputs the resolution analysis result 18. Composed. The vector set represents vector data that forms a set. However, the first vector set 28
Is composed of only one vector data.

As shown in FIG. 5, the important order determining means 3 converts the analysis result 8 into the gradation analysis result 49 and the resolution analysis result 5.
A divider 45 that divides the output into 0s and outputs, and a gradation candidate determiner 46 that obtains a hierarchy of gradations necessary for decoding with a preset encoding error or less and outputs this as a gradation candidate 51.
And a resolution candidate determiner 47 that obtains a hierarchy of resolutions necessary for decoding with a preset encoding error or less, and outputs this as a resolution candidate 52, and a hierarchy of a predetermined gradation and a hierarchy of resolutions. The rank determining unit 48 is configured to rank the gradation candidates 51 and the resolution candidates 52 as one rank and output the rank 9.

As shown in FIG. 6, the gradation / resolution information layering means 4 determines a pattern 57 for thinning out pixels based on the order 9, and a pixel thinning pattern generator 53 for outputting this.
A pixel decimator 54 that decimates and outputs pixels in the input block 7 according to a decimating pattern 57, and a signal 58 that selects a bit plane when the pixels in the input block 7 are divided into bit planes based on the rank 9 are determined. And a gradation decimator 56 that selects and outputs a bit plane according to the gradation selection signal 58.

As shown in FIG. 7, the information source coding means 5 is a multiplexer 20 for multiplexing the rank 9 which is the output of the important rank determining means 3 and the output 10 of the gradation / resolution information layering means 4. Then, the multiplexed data 12 is arithmetically coded, the coded data 11 is output if a code amount excess signal 13 described later is not output, and the arithmetic encoder 21 that terminates the encoding process when the signal 13 is output. And the data amount of the code data 11 are added, and this is compared with a preset code amount,
It is composed of a code amount adder / comparator 22 which outputs a code amount excess signal 13 which indicates that the preset code amount has been reached or is about to be reached.

Next, the operation of the above image coding apparatus will be described.

The image data 6 is sampled and divided into m × n pixel input blocks 7 composed of a plurality of pixels by the blocking means 1 shown in FIG. The output of the blocking means 1 is supplied to the gradation analyzer 14 of the gradation / resolution analysis means 2 shown in FIG.

In the tone analyzer 14, the input block 7
Is analyzed, and the resolution analyzer 15 analyzes the resolution feature of the input block 7. The gradation analysis result 17 output from the gradation analyzer 14 and the resolution analysis result 18 output from the resolution analyzer 15 are multiplexed by the multiplexer 16 and output as the gradation / resolution analysis result 8.

When the tone analyzer 14 analyzes the tone feature amount, the variance calculator 19 shown in FIG. 3 calculates the variance value of the pixels in the input block 7, and outputs this as the tone analysis result 17. To be done.

The resolution analyzer 15 corresponds to the above-mentioned Japanese Patent Application No.
The image signal analysis method proposed in the specification of No. 202129 is used. When the resolution feature quantity is analyzed in the resolution analyzer 15, as shown in FIG.
The vector set is set by the following procedure.

First, a plurality of representative shape blocks are arranged in each branch of a plurality of stages (three stages in this embodiment) of a binary tree. At this time, the representative shape block is a block in which the density gradient direction of the block is the vertical direction, the horizontal direction, the diagonal direction, or the like.
It is a block having a typical shape.

Next, the representative shape block arranged in each branch is treated as vector data, the vector difference between the two representative shape blocks arranged in a pair of branches is calculated, and the result is binary as a difference representative vector. It is placed at the node where the pair of branches of the tree split.

Finally, the sets of difference representative vectors arranged in the nodes of each stage are collected into each vector set 28, 31, 34.
Respectively stored in. The input block 7 is divided by the block divider 24 with a positive integer ratio j. Next, the average value separator 26 determines the average value of the pixels in the divided block 25, and outputs the average value separation block 36 in which this average value is subtracted from each pixel in the divided block 25. It

Next, the feature amount analysis process will be described. First, the first inner product calculator 27 causes the average value separation block 3
The inner product of 6 and the representative shape block 37 stored in the first vector set 28 is calculated, and the positive and negative signs 38 of the inner product are output. The positive / negative sign 38 of the inner product is held by the first index holder 29 as an index. Next, the second inner product calculator 30 causes the average value separation block 3
6 and the representative shape block 40 indicated by the inner index 39 of the representative shape block stored in the second vector set 31 are calculated, and the positive / negative sign 4 of the inner product is calculated.
1 is output. The positive / negative sign 41 of the inner product and the index 39 are held by the second index holder 32 as new indexes. Finally, the third inner product calculator 33 calculates the inner product of the average value separation block 36 and the representative shape block 43 indicated by the inner index 42 of the representative shape block stored in the third vector set 34, and the inner product Positive and negative signs 44 are output.
The positive and negative signs 44 and the index 42 of the inner product are held by the third index holder 35 as new indexes. Thus, the division ratio j of the input block 7
Similar processing is performed on each of the divided blocks 25 divided by
Is the third index holder 3
5 is held, and the set of held indexes is output as the resolution analysis result 18. At this time, the index held in the third index holder 35 represents the shape of the waveform when the input block 7 is regarded as a two-dimensional waveform of the image.

The resolution analysis result 18 obtained by the resolution analyzer 15 is supplied to the multiplexer 16 together with the tone analysis result 17 obtained by the tone analyzer 14 and multiplexed as shown in FIG. , Is output as an analysis result 8 from the gradation / resolution analysis means 2 shown in FIG.

Analysis result 8 from the gradation / resolution analysis means 2
Is supplied to the important order determining means shown in FIG. 5, and is distributed by the distributor 45 to the gradation analysis result 49 and the resolution analysis result 50.

The distributed gradation analysis result 49 is referred to by the gradation candidate determination unit 46, and the gradation candidate 51 is determined by the gradation candidate determination unit 46. At this time, the gradation analysis result 4
9, that is, the number of gradations is assigned according to the magnitude of the dispersion value of the pixels in the input block 7, and this is set as the gradation candidate 51. For example, the larger the dispersion value, the larger the number of gradations assigned.

The distributed resolution analysis result 50 is
The resolution candidate determiner 47 is referred to, and the resolution candidate determiner 47 determines the resolution candidate 52. At this time,
Depending on the resolution analysis result 50, that is, the waveform shape of the input block 7, a resolution corresponding to a sampling period capable of reproducing the waveform is assigned, and this is set as a resolution candidate 52. For example, when the gradation change cycle of the waveform shape of the input block 7 is long, that is, when a high spatial frequency component is not included, a long sampling cycle, that is, a low resolution is assigned to reproduce the waveform. Further, when the input block 7 has a stepwise gradation change (edge), that is, includes a high spatial frequency component, the sampling period is shortened, that is, a high resolution is assigned in order to reproduce the waveform.

The gradation candidate 51 and the resolution candidate 52 are referred to by the rank determining unit 48, and the rank determining unit 48 outputs the rank information 9. At this time, starting from a predetermined number of gradations and resolution, the order of increasing the number of gradations and resolution in order toward the gradation candidate 51 and the resolution candidate 52 is determined, and this is set as the order information 9. An example of this order determination process will be described with reference to FIGS. The horizontal axis of the graph is the pixel thinning rate, which corresponds to the resolution. The vertical axis represents the number of bits per pixel, which is Log _{2 of the} number of gradations.
Is the value obtained. The points described as candidates in the graph represent gradation candidates 51 and resolution candidates 52. First, a point described as a preset first approximation is used as a starting point. Next, from the starting point to the candidate point, the second approximation, the third approximation, the fourth approximation
Through the approximation and the fifth approximation, the rank is sequentially determined up to the candidate point. After reaching the candidate point, the order is similarly determined toward the point marked as the original image in the graph. Note that FIG.
In the above example, the first approximation point is the point with the least image information. FIG. 8 (a) shows the order determining step in the case where the slope of the straight line connecting the first approximation point and the candidate point is greater than 1, and FIG.
Shows the order determining step when the slope of the straight line connecting the first approximation point and the candidate point is smaller than 1.

The ranking information 9 from the important ranking determining means 3 is
It is supplied to the pixel thinning pattern generator 53 of the gradation / resolution information layering means 4 shown in FIG. 6, and is sequentially converted into the pixel thinning pattern 57 by the pixel thinning pattern generator 53. For example, the pixel thinning pattern 57 as shown in FIG. 9 is generated according to the thinning rate of the pixels corresponding to the resolution in the order information 9. In FIG. 9, the size of the input block 7 is 8
Pixel thinning rate and pixel thinning pattern 5 for × 8 pixels
7 shows the correspondence. Although the pixel thinning pattern shown in FIG. 9 is isotropic with respect to the two-dimensional direction of the image data, for example, as shown in FIGS. The resolution may be changed to an anisotropic pixel thinning pattern. In this case, the pixel thinning pattern is adaptively changed according to the resolution analysis result 18, that is, the waveform shape of the input block 7. For example, in the input block 7 having edges in the vertical direction, high spatial frequency components are included in the horizontal direction and high spatial frequency components are not included in the vertical direction, so that the original waveform shape is reproduced. As shown in FIG. 7A, the horizontal resolution may be high and the vertical resolution may be low. On the contrary, in the input block 7 having an edge in the horizontal direction, a pixel thinning pattern as shown in FIG.

The input block 7 is sequentially thinned by the pixel thinning unit 54 according to the pixel thinning pattern 57, and output as thinned pixels 59.

The order information 9 is also supplied to the gradation selection signal generator 55, and is converted into the gradation selection signal 58 successively by the gradation selection signal generator 55. For example, a gradation selection signal 58 for selecting a bit plane as shown in FIG. 11 is generated according to the number of bits per pixel corresponding to the number of gradations in the order information 9. FIG. 11 shows the correspondence between the number of bits per pixel and the selected bit plane when the image data 6 has a gradation of 8 bits / pixel.

The pixels 59 decimated by the pixel decimator 54 are subjected to the gradation selection signal 5 by the gradation decimator 56.
According to 8, gradations are sequentially thinned and output.

As described above, the gradation / resolution information layering means 4 passes through the candidate points from the first approximation as shown in the example of FIG. , The resolution, that is, the pixel pattern is sequentially output.

The rank information 9 from the important rank determining means 3 and the output 10 of the gradation / resolution information layering means 4 are supplied to the multiplexer 20 of the information source coding means 5 shown in FIG. As shown in the example, each time the hierarchical layer sequentially progresses from the first approximation to the original image point through the candidate point, the multiplexer 20 multiplexes and outputs. The information 12 multiplexed by the multiplexer 20 is encoded by the arithmetic encoder 21 and code data 11 is output. In this embodiment, the arithmetic coding method is used for the source coding of the multiplexed information 12, but other methods such as the Huffman coding method may be used. Code data 1 by the code amount addition / comparator 22
The sum of the data amounts of 1 is added, and the code amount excess signal 13 forcibly ending the encoding at the time when the total data amount reaches the predetermined code amount or immediately before the total data amount reaches the predetermined code amount. Is output. The arithmetic encoder 21 ends the encoding when the signal 13 is input, and stops outputting the encoded data thereafter.

With the above, all the coding processing for one input block 7 is completed, and the processing for the next input block 7 is started. This is performed for all data of the image data 6.

[0059]

As described above, according to the present invention,
The image data is divided into blocks, the gradation and resolution are analyzed for each block, and the gradation candidates and resolution candidates necessary for reproducing the original block within a predetermined coding error are obtained according to the analysis result, and the predetermined candidates are calculated. Starting from the gradation and resolution, the gradation and resolution candidates are sequentially divided into a gradation hierarchy and a resolution hierarchy, and the data of the gradation hierarchy and the resolution hierarchy are encoded and encoded. Since the encoding process is terminated when the total amount of the generated data reaches the preset code amount, 1. A constant compression ratio is always obtained for each block.

2. Local encoding / decoding is possible for each block.

3. Uniform encoding processing without complicated branch processing is performed.

4. Coding efficiency is high because it is not necessary to consider the margin for the target code amount.

5. It is not necessary to repeatedly perform encoding until the code amount control error becomes small.

6. Exception handling is unnecessary when the code amount exceeds the target value.

7. Coding efficiency is high because both visual and statistical redundancies are suppressed.

The effect is as follows.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an image encoding device of the present invention.

FIG. 2 is a configuration diagram of an embodiment of a gradation / resolution analysis means of the present invention.

FIG. 3 is a configuration diagram of a gradation analyzer of an embodiment of the gradation / resolution analysis means of the present invention.

FIG. 4 is a configuration diagram of a resolution analyzer of an embodiment of the gradation / resolution analysis means of the present invention.

FIG. 5 is a configuration diagram of an embodiment of an important rank determination means of the present invention.

FIG. 6 is a configuration diagram of an embodiment of a gradation / resolution information layering means of the present invention.

FIG. 7 is a configuration diagram of an embodiment of an encoding means of the present invention.

FIG. 8 is an operation explanatory diagram of the embodiment of the important ranking determination means of the present invention.

FIG. 9 is an operation explanatory diagram of resolution hierarchization of the embodiment of the gradation / resolution information hierarchizing means of the present invention.

FIG. 10 is an operation explanatory diagram of resolution hierarchization of the embodiment of the gradation / resolution information hierarchizing means of the present invention.

FIG. 11 is an operation explanatory diagram of gradation layering of the embodiment of the gradation / resolution information layering means of the present invention.

FIG. 12 is a configuration diagram showing an example of a conventional image coding apparatus that performs resolution approximation and gradation approximation according to the content of an image.

[Fig. 13] Fig. 13 is a configuration diagram showing an example of a conventional image encoding device that controls a code amount.

[Explanation of symbols]

1: blocking means, 2: gradation / resolution analysis means, 3:
Importance order determination means, 4: gradation / resolution information hierarchization means,
5: information source coding means, 6: image data, 7: input block, 8: analysis result, 9: rank, 10: output of gradation / resolution information layering means, 11: coded data, 12: multiplexed Data, 13: code amount excess signal 13, 14: gradation analyzer, 15: resolution analyzer, 16: multiplexer, 17:
Gradation analysis result, 18: Resolution analysis result, 19: Distributed calculator, 20: Multiplexer, 21: Arithmetic encoder, 22: Code amount adder / comparator, 24: Block divider, 25: Divided block, 26: average value separator, 27: first inner product calculator,
28: first vector set, 29: first index holder, 30: second inner product calculator, 31: second vector set, 32: second index holder, 33: third inner product calculator, 34: third Vector set, 35: third index holder, 36: average value separation block, 45: distributor, 46: gradation candidate determiner, 47: resolution candidate determiner,
48: rank determiner, 53: pixel thinning shape generator, 5
4: Pixel thinning-out device, 55: Gradation selection signal generator, 56:
Gradation decimator

Claims (4)

[Claims] 1. An image sampled from a plurality of pixels, m ×
Blocking means for dividing the input block into n pixels (m and n are positive integers), analyzing means for analyzing the resolution and gradation feature amount of the pixels in the input block, and resolution of the pixels in the input block. In the input block according to the ranking determined by the significance ranking determining means and the ranking determined by the significance ranking determining means for performing ranking of each hierarchy when divided into a hierarchy and a hierarchy of gradations. Grayscale / resolution information layering means for dividing and outputting pixels by the resolution hierarchy and the gradation hierarchy, and encoding the output of the rank and the grayscale / resolution information hierarchy means and outputting code data. An image coding apparatus comprising: a coding unit. 2. The analysis means, when analyzing the feature quantity of the resolution of the input block, has m ×
n pixels (m and n are positive integers) or their positive integer ratio j
Approximation between each of a plurality of sets of representative shape blocks made up of pixels divided by (j is a positive integer) and an average value separation block obtained by subtracting the average value in the input block from each pixel in the input block And the index of the representative shape block having the highest degree of approximation or the set of indexes of the representative shape block having the highest degree of approximation for each of j blocks is used as the feature amount of the resolution of the input block. The image coding apparatus according to claim 1, wherein 3. The encoding means adds the amount of the code data in the case of encoding the rank and the output of the gradation / resolution information layering means and outputting the code data, and the preset value is set. The image coding apparatus according to claim 1, wherein the coding process is terminated when the code amount is reached. 4. When determining the order of the hierarchy of the resolution and the hierarchy of the gradation, the importance order determining means determines
A hierarchy of a predetermined resolution and a hierarchy of gradations are set as a first rank, and a candidate of a hierarchy of resolutions and a candidate of a hierarchy of gradations that can decode the input block with a preset encoding error or less are the characteristics of the resolutions. Amount and the characteristic amount of the gradation, and obtain each independently, starting from the resolution hierarchy and gradation hierarchy of the first rank, and heading toward the resolution hierarchy candidate and the gradation hierarchy candidate. The image coding apparatus according to claim 1, wherein the image coding apparatus performs ranking according to the order.