JPH0273777A - Picture data decoder - Google Patents

Abstract

PURPOSE:To simplify the operating sequence of a picture by selecting a difference of each block from a difference decoded from a code data and a fixed value '0' based on a representative gradation number of each block detected from a resolution component in advance and writing a data by one pattern in a difference memory. CONSTITUTION:While a difference Ld decoded by a difference decoding means 2 is inputted to a difference value selection means, the mode is inputted from a mode detection means 1. A zero generating means 3 always generates '0' as a data. A difference value selection means 4 selects the difference Ld and the value '0' generated by a zero generator based on the output of the mode detection means 1, outputs the result to a difference value storage means 5 and a difference value read means 6 reads a difference value corresponding to each block from the difference value storage means 5 sequentially. Thus, the operating sequence to reproduce a picture by using the difference value is simplified and the efficiency of the operating test of a picture recovery circuit is improved.

Description

[Detailed description of the invention] [Table of contents] Overview Industrial field of application Prior art (Figures 4 and 5) Means for solving the problem to be solved by the invention (Figure 1) Example of operation (Figures 2, 3, and 4) Effects of the invention (Summary) This invention relates to an image data restoration device that restores compressed image data, and the image quality has improved step by step, especially from coarse images to high-quality images. Regarding an image data restoration device that restores an image, the purpose of the present invention is to simplify the operation sequence when reproducing an image using difference values, and to improve the efficiency of the operation test of the image reproduction circuit. Divide a multivalued image into blocks consisting of a plurality of adjacent pixels, compare the difference between the maximum and minimum gradation values within each block with a threshold value, and display the image of this block. Find multiple representative gradations, display the representative gradation as a reference value in each block, and the difference value between this reference value and the representative gradation, and calculate the resolution component that indicates the distribution of the representative gradation in each block. , an image data restoring device for restoring image data obtained by encoding the reference value, the difference value, and the resolution component, comprising: mode detection means for detecting the number of representative gradations of each block from the resolution component; The difference value decoding means for decoding, the zero generation means for always generating a zero signal as a difference value, and the mode detection means are arranged so that the number of representative gradations is 1.
difference value selection means for selecting the zero signal generated by the zero generation means, and selecting the difference value decoded by the difference value decoding means when the number of representative gradations is plural; , a difference value storage means for storing the zero signal and the difference value selected by the difference value selection means, and a difference value reading means for sequentially reading the difference value corresponding to each block from the difference value storage means. do.

(Industrial Application Field) The present invention relates to an apparatus for restoring compressed image data, and more particularly to an image data restoring apparatus that performs hierarchical restoration from a coarse image to a high-quality image. In database searches and the like, hierarchical transmission is desired in which the image quality is gradually improved from a coarse image to a high quality image so that the recipient can recognize the outline of the image from an early stage. That is, if an unnecessary image is recognized at the rough image stage, the next search can be performed without performing subsequent restoration. This is more effective because restoration of a coarse image can be performed in a shorter time than restoration of a high quality image.

(Prior Art) As a highly efficient compression method for image data, there is, for example, a multi-gradation adaptive block coding method (1986 Institute of Image Electronics Engineers National Conference Proceedings 6).

This multi-gradation adaptive block coding method (General
ized Block Truncation Cod
ing (hereinafter abbreviated as GBTC) divides the original image into blocks consisting of NXN pixels, and each pixel (X i
j ) within the maximum and minimum pixel levels within the block
(n=o.

1.2. ...) level, expresses the quantization level of each pixel in a bit plane format, and encodes gradation information and bit plane information.

This is N=4. The case where n=2 will be described in detail. FIG. 4 shows the GBTC algorithm. Each block has a maximum pixel level within the block (MAX
L) and the minimum pixel level (MINL) and the encoding parameter (threshold) T1. T2 (Tt <T2> allows the following three encoding modes (Mode A, Mode B, Mode C
)are categorized.

Mode A: If MAXL-MINL≦T1, the pixels in the block are quantized to levels (po).

Mode B: If TI<MAXL-MINL≦T2, the pixels in the block are quantized to two levels (p, , p2).

Mode C: When T2<MAXL-MINL, pixels within a block are quantized into four equally spaced levels (Q, to Q4).

The quantization level is determined by the block reference level La, level interval Ld, and level designation signals (φ, ), j, (φ
2) Described by i. If the average value processing is AVE(), each element required for encoding is calculated as follows.

Mode A: Po = AVE (Xij) = La (φx)
tj=o, (φ2), J=0 (for all 1+J) Moat B: P 1=AVE (X t J ≧(M
AX L +MINL)/2) P2=AVE (Xij< (MAXL+MINL)/
2) La= (Pt + P2 )/2 Ld=P□−P2 (φ1)ij=O: (However, if Xij ≧ (MAXL+ MINL)/2) (φx)zj=1: (However, if Xij< ( MAXL+MINL)/2) (φ2) i, +=o: (for all 1+J) Mote C: Ql = AVE (X i j≧(3MA
XL+MINL)/4) Q4=AVE (Xij< (MAXL+3MINL)
/4) La = (Ql +Q4) / 2Ld=2 (Q
, -Q4)/3 (φ1)Z, +=O, (φ2) 5j=0:
(However, in the case of Xij≧La+Ld/2) (φ□)2. , +=0. (φ2) ij=1: (However, in the case of La + L d / 2 > Xij≧La) (φ□) ij=1. (φ2) 1==O: (However,
When La>Xij>= La − L d / 2) (φ, )1j=1. (φ2) ij=1: (However, L
a − L d / 2 > Xij) The resolution components (φ□, φ2) are connected between blocks and
The images are converted into two bitmaps and encoded using the MMR encoding method, which is a standard encoding method for binary images. Ld is variable length coding after nonlinear quantization, La is DPCM
After nonlinear quantization of the pre-difference (ΔLa) using encoding,
Perform variable length encoding.

In the GBTC method, the ISO contribution ISO/TC97/SC2/W (: N510 ”G
eneralized Block Truncati
As described in ``On Coding'', a progressive build-up function can be realized using internal stages.To support this, conventionally, encoded data is transmitted sequentially from the first stage to the second stage. 3rd and 7th
It is designed to shorten the image data transmission time by transmitting image data to the stage and allowing the person using this image data to stop acquiring the image data if necessary. be.

FIG. 5 shows an image data restoration device that restores this image data. In the figure, 11 is an input terminal;
12 is a signal distribution unit that decomposes the input image data into name components (reference value La, difference value Ld, resolution components φ1 and φ2) and distributes them to each decoding unit; 13 is a reference value decoder that decodes the reference value La; 14 is a difference value decoding unit that decodes the difference value Ld, and 15 is a resolution component φ1. A resolution component decoding unit that decodes φ2 is shown. In addition, in the same figure, 16 is a resolution component storage unit that stores this decoded resolution component, and 17 is a resolution component storage unit that stores the decoded resolution component, and 17 is a resolution component storage unit that stores the decoded resolution component.
, mode C), a mode detection section 18 detects the mode signal from the mode detection section 17 and the resolution component φ1. φ
2 and the difference value signal Ld, a representative gradation value calculation unit calculates and outputs a representative gradation value, and 19 indicates an update memory that stores decoded image data.

A signal input from the terminal 11 is distributed by the signal distribution section 12 to each decoding section 13, 14, and 15, code data of the reference value La is decoded by the reference value decoding section 13, and the value of La is output. An image signal S=La is given to all pixels of the block, and this is stored in the update memory 19. When all La of one screen is decoded, a schematic image is constructed by Sl. The above restoration process for obtaining 81 is referred to as "search mode restoration."

Next, when the code data of φ is input, the resolution component decoding unit 15 decodes φ, and the decoded signal of φ□ for at least four lines is stored in the resolution component storage unit 16. and,
In the mode detection unit 17, φ of each block is “1”.
Calculate the number of gradations and detect the number of gradations. If φ1 is "0" for all pixels in one block, the block has one gradation expression, and if there is at least one pixel with φ1 "1" in one block, it has two gradation or four gradation expression. It is determined that the block is The representative gradation value calculation unit 18 calculates, based on the number of gradations in the block detected by the mode detection unit 17, the image signal S stored in the update memory 19, and the difference value Ld for each block. The representative gradation value of each pixel in the block is calculated as follows, and the contents of the update memory 19 are rewritten from Sl to 32.

(1) In the case of a block with one gradation expression: Since the difference value Ld does not exist, the difference value decoding section 14 does not decode the code data of the difference value Ld, but adds a fixed value "0" to the image signal S□.

52=S, +O ■In the case of a block with 2-gradation expression or 4-gradation expression: Since the difference value Ld exists, the code data of the difference value Ld is decoded by the difference value decoding unit 14, and the decoded difference value Ld From the value of , the variance value "±1/2LdJ is added to the image signal S1 according to the value of φ of each pixel.

The pixel with φ1=0 is 52=S□+1/2Ldφ1=1
52=St -1/2Ld These processes are repeated for one screen.

Subsequently, when the code data of φ2 is input, the resolution component decoding unit 15 decodes φ2, and the code data of φ2 for at least four lines is decoded by the resolution component decoding unit 15.
The decoded signal of is stored in the resolution component storage 16. Then, in the mode detection unit 17, “1” of φ2 of each block is detected.
” and detect the number of gradations. If φ2 is “0” for all pixels in one block, one gradation expression or two
Determined as a gradation expression block. Furthermore, if there is at least one pixel with φ2 of "1" in one block, the block is determined to be a four-gradation expression block. Representative gradation value calculation unit 1
8, based on the number of gradations in the block detected by the mode detection unit 17, the image signal S2 stored in the update memory 19, and the value of the difference value Ld for each block,
Calculate the representative tone value of each pixel in the block as follows,
The contents of the update memory 19 are rewritten from 82 to 83.

(1) In the case of a block with one gradation expression: Since the difference value Ld does not exist, the difference value decoding unit 14 does not decode the code data of the difference value Ld, but adds a fixed value "0" to the image signal S2.

53=82+0 ■For a block with two-gradation expression: Since the difference value Ld exists, the difference value decoding unit 14 decodes the code data of the difference value Ld, but the variance value with the image signal S2 is "0". Therefore, the fixed value "0" is added to the image signal S2 without using the value of the decoded difference value Ld.

S3 = S2 +0 ■For a block with 4-gradation expression: Since the difference value Ld exists, the difference value decoding unit 14 decodes the code data of the difference value Ld, and from the value of the decoded difference value Ld, the variance value is calculated. ``Add ±1/4LdJ to the image signal S2 according to the value of φ2 of each pixel.

The pixel of φ2=O is S3 = 32+1/4Ldφ2=1
The pixel of S3 = 32-1/4Ld These processes are 1
Repeat for the screen.

S3 obtained in this way represents the restored image.

The above restoration process for obtaining 83 is called "standard mode restoration."

[Problem to be solved by the invention]

By the way, in conventional image data restoration devices, when updating the update memory using the difference value, the number of gradations of each block is first detected from the resolution component, and the number of gradations of each block is determined according to the detected number of gradations. As the difference value, select whether to use a value obtained by decoding the code data of the difference value or use a fixed value "0", calculate the representative gradation value of each pixel using the selected difference value, The image was being played.

That is, since the image is reproduced using the difference value both in the layer of φ and in the layer of φ2, the above series of processing must be performed for each block for one screen. There is a problem that the operation sequence during playback becomes complicated.

In addition, since there are blocks in which no difference value exists, the difference value decoded from the encoded data by the difference value decoding means is the same as the image.
Since there is no one-to-one correspondence, there is a problem that the efficiency of testing the operation of the image reproducing circuit is low. SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide an image data restoring device that can simplify the operation sequence when reproducing an image using the difference value and improve the efficiency of the operation test of the image reproducing circuit.

(Means for determining the problem) As shown in FIG. 1, the means for solving the above problem in the present invention is to divide a multilevel image into blocks each consisting of a plurality of adjacent pixels, and For each block, compare the difference between the maximum gradation value and the minimum gradation value within the block with a threshold value to determine one or more representative gradations for displaying the image of this block,
The representative gradation is displayed as a reference value in each block and a difference value between this reference value and the representative gradation, a resolution component indicating the distribution of the representative gradation in each block is calculated, and the above reference value and difference value are calculated. An image data restoring device for restoring image data encoded with a resolution component includes a mode detection means 1 for detecting the number of representative gradations of each block from the resolution component, and a difference value decoding means 2 for decoding the difference value. When the number of representative gradations is 1, the mode detection means selects the zero signal generated by the zero generation means, and also selects the zero signal generated by the zero generation means when the number of representative gradations is 1. When there are multiple keys, a difference value selection means 4 for selecting the difference value decoded by the difference value decoding means 2, and a difference value for storing the zero signal and difference value selected by the difference value selection means 4. It comprises a storage means 5 and a difference value reading means 6 for sequentially reading the difference values corresponding to each block from the difference value storage means 5.

[Effect]

The difference value Ld decoded by the difference value decoding means 2 is input to the difference value selection means, while the mode is input from the mode detection means 1. The zero generating means 3 always generates "0" as data. The difference value selection means 4 selects the difference value Ld and "0" generated by the zero generator based on the output of the mode detection means 1.
” is selected and output to the difference value storage means 5. i.e. 1
In the case of a block with gradation expression, Ld does not exist, so the data "0" of the zero generation means 3 is selected, while in the case of multiple gradations, Ld exists and the decoded data from the difference value decoding means 2 is selected. do. The difference value reading means 6 sequentially reads the difference values corresponding to each block from the difference value storage means 5.

〔Example〕

Embodiments of the image data restoration device according to the present invention will be described below.

FIG. 2 shows an embodiment of the image data restoration device according to the present invention.

In the figure, 21 is an input terminal, 22 is a signal distribution section, 2
3 is a reference value decoding unit that decodes the reference value La, 24 is a difference value decoding unit that decodes the difference value Ld, and 28 is a resolution component φ1.
29 is a resolution component storage unit that stores the decoded resolution component; 27 is a mode detection unit that detects the mode of the image from this resolution component; 26
indicates a difference value storage section that stores difference values. In this embodiment, the difference value decoding section 24 is provided with a difference value selection section 25 that selects and outputs a difference value and a zero signal depending on the mode. As shown in FIG.
8, a zero generator 34 that always generates a zero signal, a mode determiner 35 that determines that the mode signal from the mode detection section is mode A, and a latch section 38 that receives the outputs of the zero generator 34. A multiplexer 36 is provided which selects and outputs a zero signal when the mode determiner 35 determines the mode A, and otherwise selects and outputs the difference value latched in the latch section 38.

Note that reference numeral 37 in the figure indicates a difference value decoding control unit that transmits a data write request (WRITE) to the difference value storage unit 26. In addition, the representative gradation calculation unit 30 in the figure includes a latch unit 38 that latches the difference value (including a zero signal) from the difference value storage unit 26, and a read request (R) to the difference value storage unit 26.
An image reproduction control section 39 that generates EAD) is displayed.

Next, the operation of the image data restoration device according to this embodiment will be explained.

The signal input from the terminal 21 is distributed to each decoding section by the signal distribution section 22, the code data of the reference value La is decoded by the reference value decoding section 23, and the value of La is output.

When performing "restoration of search mode", give the image signal S0 as S = La to all pixels of each block, store this in the update memory 31,
A schematic image is constructed by Sl.

"Restoration of standard mode" is performed by the following procedure.

First, when code data of φ is input, the resolution component decoding unit 28 decodes φ□, and the decoded signal of φ1 for at least four lines is stored in the resolution component storage unit 29. Then, in the mode detection unit 27, “1” of φ1 of each block is detected.
” and detect the number of gradations. If φ□ is rOJ for all pixels in one block, it becomes a block with one gradation expression, and if there is at least one pixel with φ1 “1” in one block, it is expressed as two gradation expression or four gradation expression. It is determined that the block is The multiplexer 36 of the difference value selection unit 25 selects the mode detected by the mode detection unit 27 as the difference value Ld from the zero generator 34 based on the result of the mode determination unit 35 determining the mode detected by the mode detection unit 27.
On the other hand, in the case of a block with two-gradation expression or four-gradation expression, the difference value decoding signal unit 24 selects the value of the difference value Ld decoded from the code data, and stores the lowest value in the difference value storage unit 26. Stores the difference value of 1 block line (4 lines). By repeating these processes for one screen, the difference values for one screen are stored in the difference value storage section 26.

The operation of this part will be explained in detail with reference to FIG.
The decoded difference value Ld inputted from the terminal 31 is latched by the latch section 38 and inputted to the multiplexer 36. On the other hand, the number of gradations of each block input from the terminal 32 (
mode) is input to the mode determiner 35. The mode determiner 35 determines the number of gradations of the target block according to the input mode, and notifies the multiplexer 36 of the result (MODEA). The zero generator 34 always outputs “
0" is generated. The multiplexer 36 selects the latch unit 38 according to the determination result (MODEA) of the mode determiner 35.
The difference value Ld latched in the difference value Ld and the data "0" generated by the zero generator 34 are selected and output to the difference value storage section 26.

That is, in the case of a block expressing one gradation, since Ld does not exist, data "0" of the zero generator 34 is selected. On the other hand, in the case of blocks with two-gradation and four-gradation expression, Ld exists and is data decoded from code data by the difference value decoding means, so the data latched in the latch unit 38 is selected. The difference value decoding control unit 37 is the difference value storage unit 2
A write request (WRITE) is sent to 6, and the data selected by the multiplexer 36 is written into the difference value storage section 26. By repeating this writing for one screen, the difference values Ld corresponding one-to-one to all blocks are stored in the difference value storage section 26. Next, the image reproduction control unit 39 makes a data read request (REA) to each block difference value storage unit 26.
D), and the read data is latched by the latch section 38 and held until reproduction of one block of pixels is completed. By repeating this reading for one screen, 1
Representative tone values of all pixels on the screen are calculated.

Next, the representative gradation calculation unit 30 uses the value of the difference value Ld stored in the difference value storage unit 26 and the image signal S□ stored in the update memory 31 to calculate the representative gradation of each pixel within the block. The value is calculated as follows according to the value of the resolution component φ1, and the contents of the update memory 31 are updated from S to 82.
Rewrite it to .

■For a pixel with φ1=0: 52=S, +1/2Ld However, in a block with one gradation expression, equivalently, 52=3
1+O (Ld=O) ■For pixels with φ and -〇: S2 = S□ -1/2Ld These processes are repeated for one screen.

Subsequently, when the code data of φ2 is input, the resolution component decoding unit 28 decodes φ2, and at least four lines of φ2 are input.
The decoded signal is stored in the resolution component storage section 29. Then, in the mode detection unit 27, φ2 of each block is
1" and detect the number of gradations. If φ2 is rOJ for all pixels in one block, the block is determined to be one-gradation expression or two-gradation expression. Also, φ2 is “1”
If there is at least one pixel in one block, the block is determined to be a four-gradation expression block. Difference value selection section 25
In the case of a block with one tone expression or two tone expression, “0” is selected as the difference value Ld based on the number of tone levels in the block detected by the mode detection unit 27, and the difference value storage unit 2
Update the contents of 6 to "0". On the other hand, in the case of a block with 4-gradation expression, the signal is not written to the difference value storage unit 26 and the original data is retained, so that the minimum two block lines (4 lines) are stored in the difference value storage unit 26. The difference value is stored. By repeating these processes for one screen, the difference values for one screen are stored in the difference value storage section 26.

Next, the representative gradation value calculation section 30 uses the value of the difference value Ld stored in the difference value storage section 26 and the update memory 31.
From the pixel signal S2 stored in the pixel signal S2, the representative gradation value of each pixel in the block is calculated as follows according to the value of the resolution component φ2, and the contents of the update 31 are rewritten from 82 to 83.

■For a pixel with φ2=0: 53=S2+1/4Ld However, in a block with one gradation expression or two gradation expression,
Equivalently, S 3 = S 2 + O (L d = O) ■φ1
For pixels with =0: 53=32-1/4Ld These processes are repeated for one screen.

S3 obtained through the above processing represents the restored pixel in the standard mode.

〔Effect of the invention〕

According to the present invention, the difference value of each block is selected from the value of the difference value decoded from the encoded data and the fixed value "0" based on the representative number of gradations of each block detected from the resolution component in advance, and the difference value is By writing one screen worth of data into the value memory, during image playback, it is only necessary to sequentially read the difference values of each block using the difference value memory, which not only simplifies and streamlines the image operation sequence, but also This has the effect that the operation test of the reproducing circuit can be efficiently performed.

[Brief explanation of the drawing]

Fig. 1 is a principle diagram of the present invention, Fig. 2 is a block diagram showing an embodiment of the image data restoration device according to the invention, and Fig. 3 is a difference value selection of the image data restoration device shown in Fig. 2. FIG. 4 is a block diagram showing the GBTC algorithm, and FIG. 5 is a block diagram showing a conventional image data restoration device. 1...Mode detection means 2...Difference value decoding means 3...Zero generation means 4...Difference value selection means 5...Difference value storage means 6...Difference value reading means Principle of the great invention Ko@1 Figure

Claims (1)

[Claims] A multilevel image is divided into blocks each consisting of a plurality of adjacent pixels, and the difference between the maximum tone value and the minimum tone value within each block is compared with a threshold value to determine the value of this block. Find one or more representative gradations for displaying an image, display the representative gradation as a reference value in each block and the difference value between this reference value and the representative gradation, and calculate the distribution of the representative gradation in each block. In the image data restoring device that calculates a resolution component indicated by the reference value, and restores image data obtained by encoding the reference value, the difference value, and the resolution component, the mode detection means detects the number of representative gradations of each block from the resolution component. (1), a difference value decoding means (2) for decoding the difference value; a zero generation means (3) for always generating a zero signal as a difference value; and the mode detection means assumes that the number of representative gradations is 1. difference value selection in which the zero signal generated by the zero generation means (3) is selected, and the difference value decoded by the difference value decoding means (2) is selected when the number of representative gradations is plural; means (4), a difference value storage means (5) for storing the zero signal and difference value selected by the difference value selection means (4), and a difference value corresponding to each block from the difference value storage means (5). An image data restoring device comprising: a difference value reading means (6) for sequentially calling the difference value reading means (6).