JP2887842B2 - Image data restoration method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] A multi-valued image is divided into blocks each including a plurality of pixels.
A method and apparatus for restoring coded image data after orthogonally transforming pixels in a block, and to provide an image data restoration apparatus capable of performing effective hierarchical restoration while suppressing an increase in circuit scale. The original image is divided into blocks each including a plurality of N × N pixels, and for each of the blocks, the tone values of the plurality of N × N pixels are subjected to a two-dimensional discrete cosine transform to obtain a plurality of spatial frequency distributions. A transform coefficient representing the transform coefficient, quantizing a coefficient corresponding to the spatial frequency distribution of the transform coefficient, and restoring an image from code data obtained by encoding the quantized coefficient obtained by the quantization. Decoding means for decoding data into quantized coefficients; dequantizing means for dequantizing the quantized coefficients decoded by the decoding means; and inverse DCT transform of the DCT coefficients dequantized by the dequantizing means. You Inverse DCT transforming means, image data holding means for holding the image data inverse DCT transformed by the inverse DCT transforming means, and when the image data holding means holds image data, the image data holding means An adding unit for adding the held image data and the image data subjected to the inverse DCT transformation by the inverse DCT transformation unit; and an address for storing the image data added by the addition unit in the image data holding unit. Address generating means, and the inverse DCT conversion means
When it is determined that all the converted image data in the same block is zero, the image data added by the adding means is not stored in the data holding means, and the address generated by the address generating means is set to the next block address. And an invalid block determining means for changing.

[Industrial applications]

The present invention relates to an image data device for restoring a data-compressed image, and more particularly, to an image to be divided after dividing a multi-valued image into blocks each including a plurality of pixels, orthogonally transforming the pixels in the blocks, and then encoding the image. The present invention relates to a method and an apparatus for restoring data.

[Conventional technology]

As an efficient compression method for image data, for example, there is an adaptive discrete cosine transform coding method. Adaptive Discrete Cosine Transform: A
DCT), for example, divides an image into blocks of 8 × 8 pixels, converts the image signal of each block into coefficients of a spatial frequency distribution by two-dimensional discrete cosine transform, quantizes the coefficients with a threshold adapted to vision, and obtains the obtained quantum. This is a method of encoding by using a Huffman table which statistically obtains a conversion coefficient.

FIG. 6 is a block diagram of an encoding circuit of the ADCT system.
Hereinafter, the encoding operation will be described in detail with reference to FIG.

The image is divided into blocks of 8 × 8 pixels shown in the original image signal chart of FIG.
To enter. In the two-dimensional DCT conversion unit 32, the input image signal is orthogonally transformed by DCT transformation, and is transformed into a coefficient of a section frequency distribution shown in the DCT coefficient chart of FIG. Then, it outputs the result to the linear quantization unit 33.

FIG. 7 is a block diagram of the two-dimensional DCT transform unit 32. The input image signal is first converted to a one-dimensional DCT by a one-dimensional DCT conversion unit 32-1.
Convert. Then, the transposition unit 32-2 transposes the rows and columns,
The one-dimensional DCT conversion unit 32-3 performs one-dimensional DCT conversion again, and similarly, the transposition unit 32-4 transposes rows and columns. These two one-dimensional
Two-dimensional DCT is performed by DCT. In addition, one-dimensional DCT
The conversion units 32-1 and 32-3 perform the conversion using conversion constants.

On the other hand, in the linear quantization unit 33, the input DCT coefficient is
In FIG. 2, linear quantization is performed by dividing by a value of a quantization threshold value holding unit (quantization matrix) 34 composed of quantization threshold values for general DCT coefficients.

FIG. 13 is a chart of quantized DCT coefficients (quantized coefficients), and is a result of quantization using a constant shown in FIG. 12 as an example of a threshold value for DCT coefficients. As shown in FIG. 13, the DCT coefficient below the threshold value is 0, and the DC component and a small
A quantization coefficient having a value only in the C component is generated.

The two-dimensionally arranged quantized coefficients are converted to one-dimensional by zigzag scan. FIG. 14 is an explanatory diagram of the above-described quantization coefficient scanning order (zigzag scanning). 1
When the block is set to 8 × 8, for example, dots are sequentially selected in the upper left, then the right and the lower left,. The data selectively read sequentially by the zigzag scan is input to the variable-length coding unit 35, and the variable-length coding unit 35 performs variable-length coding on the difference between the DC component at the head of each block and the DC component of the previous block. I do. For the AC component, the length (run) of the effective coefficient (coefficient whose value is not 0) (index) and the invalid coefficient (coefficient of 0) up to that value
Is variable-length coded for each block. DC and AC components are encoded using a code table 36 composed of Huffman tables created based on statistics for each image, and the obtained codes are sequentially output from a terminal 37.

On the other hand, the above-mentioned code data is restored to an image by the following method. FIG. 8 is a block diagram of a restoration circuit of the ADCT system. The code data input from the terminal 40 is input to the variable length decoding unit 41. The variable-length decoding unit 41 decodes the code data input by the decoding table 42 constituting a table opposite to the Huffman table of the above-described code table 36 into fixed-length data of an index and a run. Output.
The inverse quantization unit 43 restores the DCT coefficient by multiplying the input quantization coefficient by the value stored in the inverse quantization matrix 44,
Output to the dimensional inverse DCT transform unit 45. The two-dimensional inverse DCT transform unit 45 orthogonally transforms the input DCT coefficients by inverse DCT transform, and converts the spatial frequency distribution coefficients into image signals.

More specifically, the two-dimensional inverse DCT transform unit 45 will be described. FIG. 9 is a block diagram of a two-dimensional inverse DCT transform unit. Terminal 40
The input DCT coefficient is subjected to a dimensional inverse DCT transform by a one-dimensional inverse DCT transform unit 41-1 and output to a transpose unit 41-2. Transposed part 41
At -2, the rows and columns of the coefficients in one block are exchanged and output to the one-dimensional inverse DCT transformer 41-3. One-dimensional inverse DCT converter 41
-3 performs one-dimensional inverse DCT on the input transposed coefficients again, and outputs the result to the transposed unit 41-4. The transposed part 41-4 is the transposed part 41
Similarly to -2, the rows and columns of the coefficients in one block are exchanged again. The signal obtained by the above operation is output from the terminal 46. That is, the image is restored. The one-dimensional inverse DCT transform units 41-1 and 41-3 convert using the inverse DCT transform constant.

 The above-described decoding will be described in more detail.

FIG. 15 is a block diagram of a conventional restoration circuit (hierarchical restoration). A conventional hierarchical restoration method will be described with reference to FIG.
The code data is input from a terminal 51 to a variable length decoding unit 52.
Assuming that the scanning order of the DCT coefficients is X01, X02... X64 (corresponding to 1 to 64 in FIG. 14), the variable length decoding unit 52 first
The code data corresponding to the DCT coefficient of the n1 term (n1 <64) is decoded to the quantized DCT coefficient using the decoding table 53. The decoded quantized coefficient is input to the inverse quantization unit 54, and the inverse quantization unit 54
Performs inverse quantization to DCT coefficients using the quantization matrix 55. The inversely quantized DCT coefficients are converted into image data by a two-dimensional inverse DCT transform unit 56. The obtained image data is added to the adder 57.
Are stored in the image memory 58 via the terminal 59 and output from the terminal 59. Next, the variable length decoding unit 52 calculates the n1 to n2 terms (n1 <n2
≤ 64) D is the quantized code data corresponding to the DCT coefficient
Decode to CT coefficients. The decoded quantized coefficient is inverse quantized.
It is output to 54 and inversely quantized to DCT coefficients. The inversely quantized DCT coefficients are converted into image data by a two-dimensional inverse DCT transform unit 56. Then, the adder 57 adds the obtained image data and the image data held in the image memory 58, stores the added image data again in the image memory 58, and outputs it from the terminal 59. Finally, the DCT coefficient buffer restores the DCT coefficients from the first term to the sixty-fourth term into image data and stores them in the image memory 58. in this way,
By sequentially adding the code data and finally transferring all the data, the hierarchical restoration for one screen is completed.

[Problems to be solved by the invention]

In the prior art, when the image is updated by restoring image data (difference data) for each DCT coefficient divided into a predetermined number, the image data held in the image memory and the newly restored image data (Difference data) is added to all pixels. However, as shown in FIGS. 16 (a) and 16 (b), the high-order DCT coefficients have many zeros, and when the number of divisions increases, the DCT coefficients in the block are all restored to zero. Since all the image data becomes zero, there has been a problem that unnecessary processing is frequently performed.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an image data restoring method and an apparatus for performing effective hierarchical restoration while suppressing an increase in circuit scale.

[Means for solving the problem]

FIG. 1 is a block diagram of the present invention. The present invention divides an original image into blocks each including a plurality of N × N pixels, and performs a two-dimensional discrete cosine transform of the tone values of the plurality of N × N pixels for each of the blocks to represent a plurality of spatial frequency distributions. An apparatus for obtaining a transform coefficient, quantizing the transform coefficient into a coefficient corresponding to the spatial frequency distribution, and restoring an image from code data obtained by encoding the quantized coefficient obtained by the quantization.

The decoding means 1 decodes the input code data into quantized coefficients. This decoding is performed using a decoding table.

The inverse quantization means 2 inversely quantizes the quantized coefficient decoded by the decoding means 1. This inverse quantization is performed using an inverse quantization matrix.

The inverse DCT transform unit 3 performs an inverse DCT transform on the DCT coefficient inversely quantized by the inverse quantization unit 2.

The image data holding means 4 is, for example, the inverse DCT transform means 3
Is a memory for holding the image data subjected to the inverse DCT transformation.

When the image data is held in the image data holding unit 4, for example, the adding unit 5 selects and outputs the output of the inverse DCT transform unit 3 of the low frequency component in the first stage, and outputs the output in the second stage. The image data held in the image data holding means 4 and the image data subjected to the inverse DCT transformation by the inverse DCT transformation means 3 are added (output).

The address generating means 6 generates an address for storing the image data added by the adding means 5 in the image data holding means 4.

The invalid block judging means 7 uses the inverse DCT transforming means 3
The image data added by the adding means 5 when the image data in the block subjected to the inverse DCT conversion is determined to be zero after the first time is stored in the image data holding means 4 without being stored by the address generating means 6. Change the generated address to the next block address. For example, a counter for counting the number of image data in the block is provided, and when all the data input while the counter counts the number of image data are zero, the address is changed to the next block address.

[Action]

When the code data is added, the decoding means 1 decodes the code data using a decoding table to obtain a quantized coefficient. This quantized coefficient is inversely quantized by the inverse quantization means 2 using an inverse quantization matrix to obtain a DCT coefficient. The inverse DCT transform unit 3 performs an inverse DCT transform of the DCT coefficient inversely quantized by the inverse quantization unit 2 to obtain image data. In the first stage, the image data is stored in the image data holding unit 4 at a position indicated by the address generated by the address generation unit 6. Also,
In the second stage, the image data held in the image data holding means 4 and the image data subjected to the inverse DCT transformation by the inverse DCT transformation means 3 are added by the addition means 5 and stored in the image data holding means 4. The invalid block judging means 7 always judges whether the output of the inverse DCT transforming means 3, that is, the image data is all zero, and when all the data are zero, the adding means 5 is not operated and the block address is incremented, for example. .

In other words, the input code data is decoded into quantized coefficients, and the decoded
Is converted to T coefficients, and when one or more of the image data in the block obtained by performing the inverse DCT conversion of the DCT coefficients is not 0,
The stored image data and the image data in the block are added and stored. The input code data is decoded into quantized coefficients, and the decoded quantized coefficients are dequantized.
When the image data is converted to DCT coefficients and all the image data in the block obtained by performing the inverse DCT on the DCT coefficients are 0, the stored image data is left as it is, and the image data of the next block is decoded.

〔Example〕

FIG. 2 is an overall block diagram of an ADCT restoration section according to the embodiment of the present invention. The first-stage code data D1 (X01... Xn1: n1 <64) divided in advance from the terminal 11 is decoded by the variable-length decoding unit 12, and the inverse quantization unit 13 performs an inverse quantization process. The decoding in the variable length decoding unit 12 and the inverse quantization in the inverse quantization unit 13 are the same as those in the related art. The DCT coefficients inversely quantized by the inverse quantization unit 13 are applied to a two-dimensional inverse DCT transform unit 14, where the DCT coefficients are inverse DCT transformed and restored to image data. When the first stage, that is, the first restored image data is stored in the image memory 16, the divided image data restored by the selector 9 is selected by the signal FIRST indicating the first stage,
Write signal WRITE output from image memory control unit 10
Is written to the image memory 16. By repeating the above processing for all the blocks, the image restoration of the first stage for one screen is completed.

Next, the second stage code data D2 (Xn1 + 1... Xn2: n
1 <n2 <64) is restored to image data in the same process as above. Then, the restored divided image data is added to the invalid block detection unit 17.

FIG. 3 is a block diagram of the invalid block detection unit 17.
The 64 pixels in the signal block generated by the 6-bit counter 72 are read from the two-dimensional DCT converter 14 by the read signal DR generated by the detection controller 71, and the two-dimensional inverse DCT converter 14 Output image data (D
ATA) is zero by the comparing unit 70. In addition to the above-described image data DATA, an input serving as a reference for “0” is added to the comparison unit 70, and it is detected whether the reference for “0” is equal to the image data (DATA). Then, the comparing section 70 outputs the non-zero signal NEQ when they do not match. It is detected whether or not all pixels in one block are zero. If one block, that is, all 64 pixels are all zero, the detection control unit 71 sends the detection signal ZERO to the address generation unit 18 and the image memory control. Department
Output to 10. When ZERO is detected, the address generation circuit 18 increments the block address without accessing the image memory 16.

FIG. 4 is a block diagram of the address generation circuit 18. OR
When the detection signal ZERO is applied to the block address generation counter 82 via the circuit 81, the block address generation counter 82
Increases by one address. Then, the processing of the block ends. On the other hand, when the detection signal ZERO is not detected, the restored divided image data is added to the image data of the previous stage of the adding unit 15. And in this case the selector
19 selects this addition result.

The address generator 18 generates an address to access the image memory 16 using a 6-bit counter 80 and a block address generation counter 82, and adds the address to the image memory 16. Therefore, the above-described addition result selected by the selector 19 is stored at the same address as the address from which the stored image data was read, in accordance with the write signal WRITE output from the image memory control unit 10. At this time, the address update request signal RE output from the image memory controller 10 at the same time.
A 6-bit counter 80 for generating an address in a block is incremented by 1 by Q, and the write address ADR to the image memory 16 is updated. When the update of all the pixels in the block is completed, the carry signal CARRY is output to the 6-bit counter 80.
And the value of the block address generation counter 82 is increased by 1 via the OR circuit 81, and the processing of the block ends. By repeating the above processing for all the blocks, the image restoration of the second stage for one screen is completed.

By performing the same processing as the processing of the second stage up to the code data Di (Xni + 1... X64: ni <64) of the i-th stage, the hierarchical restoration for one screen is completed.

In this embodiment, the invalid block detection signal ZERO is two-dimensional inverse.
Although the case where it occurs in the invalid block detection unit 17 after the DCT transform has been described, it may occur in the variable length decoding unit 12, the inverse quantization unit 13, or the two-dimensional inverse DCT transform unit 14.

Further, the operation of the embodiment of the present invention will be described in detail with reference to a flowchart. FIG. 5 is an operation flowchart of the embodiment of the present invention.

When the operation of the embodiment of the present invention starts, first, the variable decoding process S1 is executed. This variable length code processing S1 is performed by the variable length decoding unit 12 in FIG. When the variable length decoding processing S1 ends, the inverse quantization processing S2 is executed. This inverse quantization processing S2
Is performed by the inverse quantization unit 13. Subsequently, inverse DCT variable processing S3 is performed. This process is performed by the two-dimensional DCT transform unit 14. After the end of the inverse DCT conversion processing S3, it is determined whether or not the image data of the block has already been restored S4. This determination is made by the signal FIRST shown in the first stage. When the signal is not restored (NO), the selector
19 selects the output of the two-dimensional inverse DCT transform unit 14, and the image memory 1
Write S5 to 6 is performed. This writing is performed under the control of the image memory control unit 10. Subsequently, it is determined whether or not all blocks have been completed (S6). When all blocks have not been completed (NO), the process is executed again by the variable length decoding process S1. Note that this determination S6 is made by the image memory control unit 10.

On the other hand, when the image data of the block is already restored in the determination S4 (YES), the inverse DCT in the block is performed.
It is determined whether all the converted data is 0 (S7). This determination is made by the invalid block detection unit 17. When it is determined in the determination S7 that all the bits are zero (YES), the invalid block detection unit 17 generates a signal (ZERO) for shifting to the head of the next block to the address generation unit 18, and the address generation unit 18 sends the signal to the image memory 16. Is moved to the beginning of the next block (S8). When it is determined in the determination S7 that all the data are not zero (NO), that is, when at least one non-zero number is included in the data, the data in the image memory and the data after the reverse change are added S9. This addition is performed by the adder 15. At this time, the selector selects the output of the adder 15, and the image data inversely transformed by the two-dimensional inverse DCT transformer 14 and the image stored in the image memory 16 are stored. The result is added to the data, and the result is added to the image memory 16.

That is, the addition is performed by the adder 15 (S9), and the data is written to the image memory 16 via the selector 19 (S5). After this writing, the determination S6 is executed, and when all the blocks are completed (YE
In S), the entire process ends.

〔The invention's effect〕

According to the present invention as described above,
When non-zero data does not exist in the block when decoding the DCT coefficients individually into image data and implementing hierarchical restoration by cumulative addition, the update process of the pixels in the block is completed only by updating the block address By letting
Hierarchical restoration can be performed efficiently.

[Brief description of the drawings]

1 is a block diagram of the present invention, FIG. 2 is an overall block diagram of an embodiment of the present invention, FIG. 3 is a block diagram of an invalid block detecting section, FIG. 4 is a block diagram of an address generating circuit, and FIG. Fig. 6 is an operation flowchart of the embodiment of the present invention; Fig. 6 is a block diagram of an encoding circuit of the ADCT system; Fig. 7 is a block diagram of a two-dimensional DCT conversion unit; Fig. 8 is a block diagram of a restoration circuit of the ADCT system; 9 is a circuit block diagram of a two-dimensional inverse DCT transform unit, FIG. 10 is a diagram showing an original image signal, FIG. 11 is a diagram showing DCT coefficients, FIG. 12 is a diagram showing threshold values for DCT coefficients, FIG. FIG. 14 is a diagram showing quantization coefficients, FIG. 14 is an explanatory diagram of a scanning order of quantization coefficients, FIG. 15 is a block diagram of a conventional ADCT hierarchical restoration unit, and FIG. 16 (a) is an example of dividing DCT coefficients ( First Stage) FIG. 16B shows an example of dividing the DCT coefficients (second stage). 1 ... Decoding means, 2 ... Inverse quantization means, 3 ... Inverse DCT conversion means, 4 ... Image data holding means, 5 ... Addition means, 6
... Address generation means, 7... Invalid block determination means.

Claims (5)

(57) [Claims] An original image is divided into blocks each including a plurality of N × N pixels, and for each of the blocks, tone values of the plurality of N × N pixels are subjected to a two-dimensional discrete cosine transform to obtain a plurality of spatial frequency distributions. In a device for obtaining a transform coefficient represented by, quantizing a coefficient corresponding to the spatial frequency distribution of the transform coefficient, and restoring an image from code data obtained by coding the quantized coefficient obtained by the quantization, Decoding means (1) for decoding code data into quantized coefficients; dequantizing means (2) for dequantizing the quantized coefficients decoded by the decoding means (1); and dequantizing means (2) ) Inversely quantized DCT coefficients by inverse D
An inverse DCT transforming means (3) for performing a CT transform, an image data retaining means (4) for retaining the image data which has been subjected to the inverse DCT transform by the inverse DCT transforming means (3), and an image stored in the image data retaining means (4). When the data is held, the image data held by the image data holding means (4) and the image data inversely transformed by the inverse DCT transformation means (3) are added, and the image data holding means is added. (4) an adding means (5) for storing the image data added by the adding means (5), and an address generating means (6) for generating an address when storing the image data in the image data holding means (4). When the image data in the same block subjected to the inverse DCT transformation for the second time and thereafter is determined to be all zero by the inverse DCT transformation means (3), the image data added by the addition means (5) is held in the image data. Means is not stored in the means (4). Image data decompression apparatus, characterized in that the address generated from less generating means (6) and an invalid Borokku judging means (7) to change to the next block address. 2. An adder for adding an output of the inverse DCT transforming means and an output of the image data holding means, and an adder for low frequency components in a first stage. Selection means for selecting an output of the inverse DCT transform means (3) and selecting an output of the adder which is a frequency component area next to the first stage in a second stage. 2. The image data restoration device according to 1. 3. The invalid block judging means (7) has a counter for counting the number of image data in the block, and when all data inputted while the counter counts the number of image data are zero. 2. The image data restoration apparatus according to claim 1, wherein the image data is changed to the next block address. 4. An original image is divided into blocks each including a plurality of N × N pixels, and for each of the blocks, a gradation value of the plurality of N × N pixels is subjected to a two-dimensional discrete cosine transform to obtain a plurality of spatial frequency distributions. In a method of obtaining a transform coefficient represented by, quantizing the transform coefficient by a coefficient corresponding to the spatial frequency distribution, and restoring an image from encoded data obtained by encoding the quantized coefficient obtained by the quantization, Of the predetermined frequency component area, the code data is decoded into quantized coefficients, the decoded quantized coefficients are inversely quantized and converted into DCT coefficients, and the restored DCT coefficients are obtained by inverse DCT transform. When one or more of the data is not 0, the image data and the restored image data are added to the frequency component area stored in the image data holding unit as the restoration result and restored just before the frequency component area. , The image Image data restoration method characterized by storing in the same area of over data holding means. 5. An original image is divided into blocks each including a plurality of N × N pixels, and tone values of the plurality of N × N pixels are subjected to a two-dimensional discrete cosine transform for each of the blocks to obtain a plurality of spatial frequency distributions. In a method of obtaining a transform coefficient represented by, quantizing a coefficient corresponding to the spatial frequency distribution of the transform coefficient, and restoring an image from code data obtained by coding the quantized coefficient obtained by the quantization, Of the predetermined frequency component area, the code data is decoded into quantized coefficients, the decoded quantized coefficients are inversely quantized and converted into DCT coefficients, and the restored DCT coefficients are obtained by inverse DCT transform. An image data restoring method, wherein when all the data is 0, the code data of the next block is immediately decoded.