JP2624900B2 - Image data restoration device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data restoration apparatus which divides a multi-valued image into blocks each consisting of a plurality of pixels, and restores an image from encoded code data after orthogonally transforming the pixels in the blocks. The present invention relates to an image data restoration device that divides one block of code data and restores an image hierarchically over a plurality of stages.

In order to store image data whose information amount is orders of magnitude larger than numerical data, in particular, data of halftone images and color images, or to transmit them at high speed and with high quality, it is necessary to use a pixel-by-pixel level. The key values need to be encoded efficiently.

A highly efficient compression method for image data is as follows.
For example, adaptive discrete cosine transform coding (Adaptive Dis
The Crete Cosine Transform (hereinafter abbreviated as "ADCT") is known.

In the ADCT method, an image is divided into blocks each having 8 × 8 pixels, and an image signal of each block is converted into DCT coefficients of a spatial frequency distribution by a two-dimensional discrete cosine transform (hereinafter, referred to as “DCT transform”). Then, the quantized DCT coefficients are quantized by a threshold value suitable for the visual sense, and the obtained quantized DCT coefficients are encoded by a Huffman table statistically obtained.

[0005]

2. Description of the Related Art FIG. 15 is a block diagram of a conventional ADCT encoding circuit. In FIG. 15, the image is divided into blocks of 8 × 8 pixels shown in FIG.
It is input to the dimensional DCT transform unit 52. In the two-dimensional DCT transform unit 52, the input image signal is orthogonally transformed by DCT, converted into DCT coefficients having a spatial frequency distribution shown in FIG. 18, and output to the linear quantization unit 54. Specifically, FIG.
As shown in (1), the image signal input from the terminal 64 is one-dimensionally DCT transformed by the one-dimensional DCT transformation part 66,
8 is output. The transpose unit 68 interchanges the rows and columns of the coefficients in one block and outputs the result to the one-dimensional DCT transform unit 70. One-dimensional DCT unit 70 performs one-dimensional DCT on the input transposed coefficients again, and outputs the resultant to transposed unit 72.
The transposition unit 72 outputs a signal obtained by exchanging the rows and columns of the coefficients in one block again to the linear quantization unit 54 from the terminal 74, similarly to the transposition unit 68.

Referring again to FIG. 15, in the linear quantization section 54, the input DCT coefficient is quantized to each of the quantization matrices 56 constituted by the threshold values shown in FIG. The control parameter is multiplied to obtain a quantization threshold, and a DCT coefficient is divided by the quantization threshold to perform a linear quantization for obtaining a quantized DCT coefficient. As a result of this quantization, as shown in FIG. 20, the DCT coefficient equal to or smaller than the threshold value becomes zero, and a quantized DCT coefficient having only the DC component and a small AC component has a value.

[0007] The quantized DCT coefficients arranged two-dimensionally are converted into one dimension by zigzag scan shown in FIG. 21 and input to the variable length coding unit 58. The variable length coding unit 58 calculates the DC component at the head of each block and the D component of the previous block.
The difference from the C component is variable-length coded. For the AC component, a variable-length coding is performed on the value of a non-zero effective coefficient (hereinafter, referred to as an “index”) and the length of a run of an invalid coefficient that becomes zero up to the index (hereinafter, referred to as a “run”). Further, the DC and AC components are encoded using a code table 60 constituting a Huffman table created based on the statistics for each image, and the obtained code data is sequentially output from a terminal 62. Is done.

FIG. 22 shows a block diagram of a conventional ADCT restoration circuit. In FIG. 22, code data input from a terminal 10 is input to a variable length decoding unit 12. The variable length decoding unit 12 decodes the input code data into fixed-length data of an index and a run by using a decoding table 14 composed of a table opposite to the Huffman table of the code table 60 in FIG. Output to the inverse quantization unit 16. The inverse quantization unit 16 inversely quantizes the input quantized DCT coefficients by multiplying each of the quantization matrices 18 by a value (quantization threshold value) multiplied by a quantization control parameter to obtain a D value.
The CT coefficients are restored and output to the two-dimensional inverse DCT transform unit 20.

[0009] The two-dimensional inverse DCT transform unit 20 orthogonally transforms the input DCT coefficients by inverse DCT transform, and converts the DCT coefficients of the spatial frequency distribution into image signals. In particular,
As shown in FIG. 23, the DCT coefficient input from the terminal 78 is one-dimensional inverse DCT transformed by the one-dimensional inverse DCT transform unit 80 and output to the transposition unit 82. The transposition unit 82 interchanges the rows and columns of the coefficients in one block and outputs the result to the one-dimensional inverse DCT transform unit 84. One-dimensional inverse DCT unit 84 performs one-dimensional inverse DCT on the input transposed coefficients again and outputs the result to transposition unit 86. The transposition unit 86 outputs a signal obtained by exchanging the rows and columns of the coefficients in one block again from the terminal 88 in the same manner as the transposition unit 82, so that the image is restored.

By the way, even if a gradation image or a color image is compressed at a high efficiency, it is difficult to transfer image data of one screen.
It takes about 10 seconds with SDN. Therefore, it takes a considerable time until the entire screen is displayed on the display device, which is inconvenient.

Therefore, the code data is stored in a plurality of layers (stages).
The image data is restored and output little by little for each layer, and a rough whole image is displayed on the display device in a short time, and thereafter, detailed image data is gradually restored to gradually display a high-quality image. Has been proposed.

FIG. 24 shows a block diagram of a conventional ADCT hierarchy restoration circuit. In FIG. 24, the code data is
0 is input to the variable length decoding unit 12. Variable length decoding unit 1
In step 2, code data is decoded into quantized DCT coefficients. The decoded quantized DCT coefficients are output to the inverse quantization unit 20 and inversely quantized to DCT coefficients. The DCT coefficients are once stored at 90 in the DCT coefficient buffer for reordering.

Here, 64 DCT coefficients are converted into X01,
X02,..., X64, the DCT coefficient buffer 90
Outputs the data from the first term to the n ₁ -th term (n ₁ <64) to the two-dimensional DCT transform unit 20, and inversely transforms the data into image data. The obtained image data is held in the image memory 92 and output from the terminal 36.

[0014] Next, DCT coefficient buffer 90 is outputted to the n ₁ first n ₂ paragraphs (n ₁ <n ₂ ≦ 64) until the data two-dimensional inverse DCT unit 20, inverse transform to the image data . Then, the obtained image data is stored in the image memory 92 and output from the terminal 36.

Finally, the DCT coefficient buffer 90 outputs all the DCT coefficients from the first term to the sixty-fourth term to the two-dimensional inverse DCT transform unit 20, restores the data to image data, and
To hold.

As described above, by dividing the data and transferring the data, and finally transferring all the data, the hierarchical restoration for one screen is completed.

[0017]

However, in such conventional ADCT gradation restoration, for example, when 8-bit (256 gradation) image data is used, the DCT coefficient after inverse quantization is calculated. An 11-bit DCT coefficient buffer 90 to be held and an 8-bit restored image memory 92 are required. Therefore, it is necessary to have a 19-bit image memory as a whole, which increases the size of the apparatus and requires processing time for memory access. However, there is a problem that the processing efficiency of hierarchical restoration is low.

The present invention has been made in view of such conventional problems, and has as its object to provide an image data restoration apparatus capable of performing efficient hierarchical restoration while suppressing an increase in circuit scale.

[0019]

FIG. 1 is a diagram illustrating the principle of the present invention. 12 is a decoding means for decoding divided code data obtained by dividing one block into a plurality of layers into quantized DCT coefficients;
16 dequantizes the decoded quantized DCT coefficient to obtain a DC
Inverse quantization means for obtaining T coefficients; 20, inverse DCT conversion means for performing inverse DCT conversion of DCT coefficients by matrix operation to obtain N + 2 bit divided image data; 22, cumulative image data and image data from inverse DCT conversion means 20; Is an image data holding means using an image memory having a size of N + 1 bits in the depth direction for storing the accumulated image data.

Numeral 30 denotes N obtained by the inverse DCT transform means 20
Level shift means for generating a divided image data without underflow by adding a value obtained by raising 2 to the power of (N-1) to the +2 bit divided image data;
4 and the N + 1-bit accumulated image data g (i-1) up to the (i-1) th hierarchy and the level shift means 30
Is a cumulative calculation means for calculating the cumulative image data g (i) up to the i-th hierarchy by using the divided image data f (i) 'of the i-th hierarchy output from.

[0021]

The divided code data obtained by dividing one block into a plurality of hierarchies is input to a decoding means, which decodes the divided code data into quantized DCT coefficients, and which is decoded by an inverse quantization means. DCT coefficient is inversely quantized to DC
A T coefficient is generated, and the DCT is
The coefficients are subjected to inverse DCT transform by matrix operation to obtain N + 2 bit divided image data. The adding means 22 adds the accumulated image data of N + 1 bits up to the (i-1) th hierarchy held in the image data holding means 24 and the divided image data in the ith hierarchy output from the inverse DCT transforming means 20. The accumulated image data up to the i-th layer is obtained and stored in the image data holding unit 24.

In the level shift means 30, the inverse D
The divided image data f output from the CT conversion means 20
(I) is added with a value obtained by raising 2 to the power of (N-1), and
Generates divided image data f (i) 'with no flow,
In the accumulating operation means 32, the accumulated image data g up to the (i-1) th hierarchy held in the image data holding means 24
(I-1) and the i-th output from the level shift means 30
The following equation g (1) = g (0) + f (1) '(where i = 1) using the hierarchical divided image data f (i)'

[0023]

(Equation 2)

Thus, the accumulated image data g up to the i-th layer
(I) is calculated and stored in the image data holding means 24.

As described above, the divided image data is sequentially restored from the divided code data obtained by dividing one block into a plurality of hierarchies, and the divided image data is cumulatively added and held, whereby the image is hierarchically restored. Therefore, the DCT coefficient buffer required in the conventional method is not required, and the apparatus scale can be reduced, the time required for memory access can be reduced, and the processing efficiency of hierarchical restoration can be improved. Also, since the accumulated image data is extended by one bit and the underflow is taken into account as (N + 1) bits, a highly accurate image restoration can be performed. Further, since the underflow is eliminated by level shifting and the overflow can be expressed accordingly, both the overflow and the underflow can be considered even with (N + 1) bits. High-precision image restoration is possible.

[0026]

FIG. 2 is a block diagram showing an embodiment of the present invention. In FIG. 2, the variable length decoding unit 12, the decoding table 1
4. Linear inverse quantization unit 16, quantization matrices 18 and 2
The dimensional inverse DCT transform unit 20 is the same as the conventional device shown in FIG. However, in the present invention, the two-dimensional inverse DCT
Following the conversion unit 20, an adder 22 and an image memory 24 as image data holding means having at least N + 1 bits in the depth direction are provided for hierarchical restoration. Adder 22
When the image data restored from the divided code data is stored in the image memory 24, the image data already stored in the image memory 24 is read, added and stored.

Here, since the gradation value of the pixel of the original image to which the present invention is applied is represented by N = 8 bits, the image memory 24 has a configuration of N + 1 = 9 bits.
That is, it is possible to store the gradation values 0 to 384 obtained by adding the gradation values for the extended 1 bit to the gradation values 0 to 256 corresponding to N = 8 bits.

The comparison unit 26 limits the bits of the image data read from the image memory 24 and outputs the image data from a terminal 28 to a display device or the like. That is, since the image memory 24 has N + 1 = 9 bits extended by 1 bit, the overflow of the image data whose gradation value exceeds 255 is fixed to 255, while the gradation value is negative. An underflow having a value is fixed to 0.

Next, the hierarchical restoration operation according to the embodiment of FIG. 2 will be described. Now, the code data D1 (X01, X02,..., X) of the first hierarchy (stage) divided in advance from the terminal 10
n ₁ , where n ₁ <64, is input, and the code data D 1 of the first stage is decoded by the variable length decoding unit 12 using the decoding table 14, and is supplied to the linear inverse quantization unit 16.
The linear inverse quantization unit 16 performs an inverse quantization operation of multiplying the input DCT coefficient by a quantization threshold obtained from the quantization matrix 18 to restore the DCT coefficient. The DCT coefficient restored by the linear inverse quantization unit 16 is a two-dimensional inverse DCT transform unit 2
0 and restored to image data. 2D inverse DC
The divided image data of the first stage restored by the T conversion unit 20 is supplied to an adder 22 and added to the already stored accumulated image data read from the image memory 24 up to the previous stage. Is output to the image memory 24. The image memory 24 holds the divided image data input by the cumulative addition, and the comparing unit 26 performs the overflow and the underflow.
After removing the flow, the signal is output from the terminal 28.

Next, the second stage code data D2 (Xn ₁
+1... Xn ₂ , where n ₁ <n ₂ ≦ 64)
The image data is restored by the same processing as the image data, added to the accumulated image data up to the previous stage by the adder 22, and stored in the image memory 24.

In this manner, by performing the processing up to the last image data X64 in the last stage, the hierarchical restoration for one screen is completed.

More specifically, the AD of the image data shown in FIG.
In the case where the code data obtained by the CT encoding is displayed by two-stage hierarchical restoration, the DCT coefficient shown in FIG.
The divided DCT coefficients are restored as shown in FIG.
Of course, this division is not performed on the DCT coefficients, but is realized by dividing the code data for the terminal 10 as shown in FIG.

The divided DCT coefficients of the first stage shown in FIG. 5 have DCT coefficients X01 to X03 up to the third order, and the second stage of FIG.
The divided DCT coefficients of the stage have the third and subsequent orders X04 to X64. The divided DCT coefficients in the first stage in FIG. 5 are restored to the divided image data in FIG. Next, the divided DCT coefficients of the second stage shown in FIG. 6 are subjected to inverse DCT transform, and the obtained divided image data shown in FIG. Adder 22
Then, the divided image data shown in FIG. 7 and the divided image data shown in FIG. 8 are added and output to the image memory 24 for storage.

In the divided image data shown in FIGS. 7 and 8, a shaded portion indicates an underflow portion having a negative value, and the value of the underflow portion indicates that the image memory 24 has 1 bit. By expanding the data to N + 1 = 9 bits, the data can be stored, and highly accurate image restoration can be performed. Of course, for the underflow portion in the image memory 24 before the final division stage, the underflow portion is fixed to the gradation value 0 by the comparing portion 26 and displayed.

In the embodiment shown in FIG. 2, a 1-bit extended image memory having N + 1 = 9 bits is used as the image memory 24.
As the image memory 24, a memory having a configuration of N + 2 = 10 bits expanded by 2 bits may be used, and the expanded 2 bits may be used for underflow information and underflow information, respectively. That is, by using the extended 1 bit for the overflow information, the gray scale 0 to 256 of N = 8 bits becomes 128 to 0 to 384 gray scales. On the other hand, the 1 bit extended to the underflow is used. Is assigned, the gradation can be expressed in the range of 0 to -128, and as a result, the gradation can be stored over the range of -128 to 384 in the 512 gradation.

In the first embodiment described above, if the image memory 24 has a configuration of N + 2 = 10 bits, image data can be stored over a range of 512 tones from -128 to 384, and higher-precision restoration is possible. Become. However, the memory capacity becomes relatively large.

FIG. 9 is a block diagram of the second embodiment in which image data can be stored over a range of 512 tones from -128 to 384 even when the image memory 24 has an N + 1 = 9 bit configuration. In this embodiment, the image memory 24 has a configuration of N + 1 = 9 bits which is extended by 1 bit, and the underflow is eliminated by the offset so that the cumulative addition of the image data is performed.

In FIG. 9, the configuration from the variable length decoding unit 12 to the two-dimensional inverse DCT transform unit 20 is the same as that of the embodiment of FIG. A level shift unit 30 is provided following the two-dimensional inverse DCT transform unit 20, and 2 (N-1) = 128 is added to the image data obtained from the two-dimensional inverse DCT transform unit 20 to under-run. I try to eliminate the flow.

The image data that has been level-shifted by the level shift unit 30 is supplied to an accumulation operation unit 32, and the accumulation operation unit 32 performs an accumulation operation with the accumulated image data up to the previous stage stored in the image memory 24. The accumulator 32 includes a subtractor 32a and a switching circuit 32, as will be described later.
b and an adder 32c. The accumulated image data on which the accumulation operation has been performed by the accumulation operation section is stored in the image memory 24. The image memory 24 has an N + 1 = 9 bit configuration extended by one bit. A subtractor 40 is provided following the image memory 24. In the subtractor 40, 2 (N-1) = 128, which has been level-shifted by the level shift unit 30, is subtracted and output to the terminal 28 as N = 8 bits. ing.

FIG. 10 is a model explanatory diagram for analyzing the operation principle of the accumulation operation unit 32 when the image memory 24 is composed of N + 1 = 9 bits. 10, first, the block A side is a two-dimensional inverse DCT transform unit 20 and a level shift unit 30, and the block B side is a cumulative operation unit 32 and an image memory 24. Also, the image data f (i), f
(I) 'and g (i-1) are defined as follows.

F (i): result of inverse DCT transform of i stage (−128 ≦ f (i) ≦ 383) f (i) ′: data obtained by converting f (i) into a positive number (0 ≦ f ( i) '≤511) g (i-1): Contents of image memory after i-1 stage (cumulative image data) (0≤g (i-1) ≤511) Also, the bits of each part are two-dimensional inverse. The DCT transform unit 20 has N + 2 = 10 bits because it includes an overflow and an underflow, and all other bits have N + 1 = 9 bits extended by one bit.

Now, assuming that image data f (i) is obtained from the two-dimensional inverse DCT transform unit 20 at the i-th stage.
Underflow is removed by the level shift by the level shift unit 30, and the level shift unit 30 outputs image data of f (i) '= f (i) +128 (1). This level shift is as shown in FIG.

In FIG. 11, N = 8 bits indicate a gradation value of 0.
The image data having a gradation value of a total of 10 bits of underflow-1 bit and overflow-1 bit is output from the two-dimensional inverse DCT transform unit 20. By adding 2 (N-1) = 128 to the gradation value of the image data including the overflow and the underflow by the level shift unit 30, as shown in the right side of FIG. Can be converted to a positive number to a gradation value of 0 to 511. That is, −128 of the output values of the two-dimensional inverse DCT
383 is converted to a positive number from 0 to 511.

Here, the data g (0) read from the image memory 24 before processing, that is, at the initial stage, is g (0) = 0.

In this initial state, i = 1 in the first stage.
Image data f (i) = f from the dimensional inverse DCT transform unit 20
Assuming that (1) is obtained, f (i) ′ = f (1) ′ by the level shift by the level shift unit 30 and is given to the accumulation operation unit 32. At the same time, since the accumulated image data g (i-1) = g (0) up to the previous time is read from the image memory 24, the output g (i) = g of the accumulation operation unit 32 is obtained.
(1) Becomes That is, in the first stage, the level-shifted divided image data is stored in the image memory 24 as it is.

After the second stage, the level shift unit 3
Assuming that there is no 0, the accumulator 32 in the i-th stage
The output image data g (i + 1) can be expressed as g (i + 1) = g (i) + f (i + 1) (3). However, since the level shift unit 30 exists, f (i + 1) '= f (i + 1) +128 f (i + 1) = f (i + 1)'-128 (4)

Therefore, when f (i + 1) in the expression (4) is substituted into the expression (3), the accumulated image data g (i + 1) from the accumulation operation unit 32 obtained in the i (≧ 2) th stage can be obtained. ) Can be expressed as g (i + 1) = [g (i) + {f (i + 1) ′ − 128}] (5). Equation (5) shows the contents of the cumulative operation of the cumulative operation unit 32 in any stage after the second stage. Therefore, the cumulative operation unit 3 shown in FIG.
12 is a subtractor 32a for subtracting 128 from the divided image data f (i) 'from the level shift unit 30 as shown in FIG. 12A, and a divided image from the level shift unit 30 in the first stage. A switching circuit 32b for outputting data, and for outputting the image data from the subtractor 32a in the second stage and thereafter.
And the adder 32c for adding the divided image data from the switching circuit 32a and the accumulated image data g (i-1) stored in the image memory 24.

If the equation (5) is transformed into g (i + 1) = [{g (i) −128} + f (i + 1) ′] (6), the accumulative calculation unit 32 obtains the result shown in FIG. ), The accumulated image data g (i-1) from the image memory 24.
Subtractor 128d from the image memory 24 in the first stage.
0), and after the second stage, a switching circuit 32e for outputting image data from the subtractor 32d, and a switching circuit 3
An adder 32f for adding the accumulated image data from 2e and the divided image data f (i) 'from the level shift unit 30 can also be used.

Referring specifically to FIG. 9, when the original image data is N = 8 bits (256 gradations), the two-dimensional inverse D
The divided image data subjected to the inverse DCT transformation by the CT transformation unit 20 is raised to the power of 2 (N−1) = 2 ⁷ = 128 by the level shift unit 30.
Is added, and as a result of this level shift, the image data of -128 to 383 is converted to a positive number of 0 to 511 as shown in FIG. 11, and the underflow is eliminated. That is, N + 1 =
With 9 bits, underflows from -1 to -128 and overflows from 256 to 383 can be expressed, and high-precision image restoration becomes possible.

FIGS. 13 and 14 show the case where each of the divided image data shown in FIGS.
28 shows divided image data when 28 is added. As a result of the addition of the gradation value 128, all the underflow portions shown by the shaded portions in FIGS. 7 and 8 are eliminated, and conversely, the overflow portions also shown by the shaded portions on the higher order side in FIG. It can be only-part. As described above, the divided image data without the underflow portion is stored in the image memory 24 in the first stage in the first stage, and the next second stage is stored in the image memory 24 in the first stage.
The divided image data shown in FIG. 14 and the accumulation operation in the accumulation operation unit 32 are stored in the image memory 24.

Regarding the display of the image data read from the image memory 24, the subtractor 40 subtracts the value of 128 raised to the power of (N-1) = 128, returns the original 8-bit data, and outputs it from the terminal 28. Become like

The present invention has been described with reference to the embodiments.
The present invention can be variously modified in accordance with the gist of the present invention described in the claims, and the present invention does not exclude these.

[0053]

As described above, according to the present invention, a divided image data is sequentially restored from divided code data obtained by dividing one block into a plurality of hierarchies, and the divided image data is cumulatively added and held, whereby the image is hierarchically restored. With such a configuration, the DCT coefficient buffer required in the conventional method is not required, so that the apparatus scale can be reduced, the time required for memory access can be reduced, and the processing efficiency of hierarchical restoration can be improved.

Further, according to the present invention, since the accumulated image memory is (N + 1) bits and the underflow is also taken into account, a highly accurate image restoration can be performed. Further, according to the present invention, if the accumulated image memory is (N + 2) bits, the overflow can be considered in addition to the underflow, and the image can be restored with high accuracy.

Also, since the DCT coefficient is level-shifted to eliminate the underflow and express the overflow by that amount, even if the image memory has the (N + 1) -bit configuration, it can be added to the underflow. Thus, an overflow can be considered, and a highly accurate image can be restored with a small number of bits.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a configuration diagram of a first embodiment of the present invention.

FIG. 3 is an explanatory diagram of original image data restored in the present invention.

FIG. 4 is an explanatory diagram of DCT coefficients restored in the present invention.

FIG. 5 is an explanatory diagram of a divided DCT coefficient of the first stage of the present invention.

FIG. 6 is an explanatory diagram of divided DCT coefficients of the second stage of the present invention.

FIG. 7 is an explanatory diagram of divided image data when the DCT coefficients of FIG. 5 are subjected to inverse DCT transform.

FIG. 8 is an explanatory diagram of divided image data when the DCT coefficients of FIG. 6 are subjected to inverse DCT transform.

FIG. 9 is a configuration diagram of a second embodiment of the present invention.

FIG. 10 is an explanatory diagram of the model of FIG. 9;

FIG. 11 is an explanatory diagram of a bit configuration and a gradation level according to the present invention.

FIG. 12 is a diagram illustrating a first example of the accumulation operation unit based on the model of FIG.
It is a 2nd block diagram.

FIG. 13 is an explanatory diagram of image data in a case where 128 is added to the divided image data of FIG. 7 to eliminate underflow.

FIG. 14 is an explanatory diagram of image data in a case where 128 is added to the divided image data of FIG. 8 to eliminate underflow.

FIG. 15 is a block diagram of a conventional ADCT encoding circuit.

FIG. 16 is a block diagram of the two-dimensional DCT transform unit of FIG.

FIG. 17 is an explanatory diagram of an original image signal.

FIG. 18 is an explanatory diagram of DCT coefficients.

FIG. 19 is an explanatory diagram of a threshold value for a DCT coefficient.

FIG. 20 is an explanatory diagram of quantized DCT coefficients.

FIG. 21 is an explanatory diagram of a scanning order at the time of quantization.

FIG. 22 is a block diagram of a conventional ADCT restoration circuit.

FIG. 23 is a block diagram of a two-dimensional inverse DCT transform unit of FIG. 22;

FIG. 24 is a configuration diagram of a conventional ADCT hierarchy restoration circuit.

[Explanation of symbols]

 12 decoding means 16 inverse quantization means 20 inverse DCT transform means 22 addition means 24 image data holding means 30 level shift means 32 cumulative operation means

Claims (5)

(57) [Claims] 1. For each block obtained by dividing an original image into a plurality of blocks each having a plurality of pixels having N-bit gradation values, the gradation values of the plurality of pixels in the block are set to 2 In an image data restoration apparatus that quantizes DCT coefficients obtained by performing a dimensional discrete cosine transform and restores an image from code data obtained by encoding the obtained quantized DCT coefficients, the divided code data obtained by dividing one block is quantized. DC
Decoding means (12) for decoding into T coefficients, inverse quantization means (16) for inversely quantizing the quantized DCT coefficients decoded by said decoding means (12) to obtain DCT coefficients, and said inverse quantization means ( Inverse DCT conversion means (20) for obtaining divided image data by performing inverse DCT transformation on the DCT coefficients obtained from 16) by matrix operation, and the inverse DCT transformation means (2)
Image data holding means (24) using an image memory having a size of at least N + 1 bits in the depth direction for holding the image data from 0) and the inverse DCT transform means (24) in the image data holding means (24). When inputting the image data from (20), the image data holding means (24)
And an adding means (22) for reading out the image data stored in the memory and adding and storing the image data from the inverse DCT transforming means (20). The divided image data sequentially restored from the divided code data is provided. An image data restoring apparatus characterized in that an image is hierarchically restored by accumulatively adding and holding. 2. The image data restoring device according to claim 1, wherein said image data holding means (24) has a depth of 2 mm.
An image data restoring apparatus characterized in that a memory having a size of N + 2 bits obtained by bit expansion is used, and underflow information and overflow information of image data are stored in each of the expanded 2 bits. . 3. For each block obtained by dividing the original image into a plurality of blocks each having a plurality of pixels having N-bit gradation values, the gradation values of the plurality of pixels in the block are set to 2 In an image data restoration apparatus for quantizing DCT coefficients obtained by performing a dimensional discrete cosine transform and restoring an image from code data obtained by coding the obtained quantized DCT coefficients, one block of code data is divided into a plurality of layers. Split,
Decoding means (12) for decoding the code data of each layer into quantized DCT coefficients; and inverse quantization means (16) for inversely quantizing the quantized DCT coefficients decoded by said decoding means (12) to obtain DCT coefficients And inverse DCT transforming the DCT coefficient obtained from the inverse quantization means (16) by matrix operation to obtain the divided image data f (i) of the ith layer
T transform means (20) and the inverse DCT transform means (20)
Is added to the divided image data of the i-th layer obtained by
1) a level shift means (30) for adding divided values to generate divided image data without underflow;
An image data holding unit (24) having a size of N + 1 bits in the depth direction for holding the accumulated image data obtained by accumulating the divided image data of each hierarchy, and the (i) held in the image data holding unit. -1) Assuming that the accumulated image data up to the hierarchy is g (i-1) and the divided image data of the i-th hierarchy output from the level shift means is f (i) ', the accumulated image data g up to the i-th hierarchy G (1) = g (0) + f (1) ′ (where i = 1) And accumulating the divided image data sequentially restored from the divided code data of each layer and storing the divided image data. An image data restoring apparatus characterized in that an image is hierarchically restored by the following method. 4. The image data restoration device according to claim 3, wherein said accumulating means subtracts a value obtained by raising 2 to the power of (N-1) from the image data from said level shift means. Subtracting means (32a) for outputting divided image data from the level shift means (30) for the divided image data of the first hierarchy, and outputting the divided image data for the divided image data of the second hierarchy and thereafter from the subtracting means (32a). Divided image data switching means (32b) for outputting divided image data
Adding means (32) for adding the divided image data from the divided image data switching means (32b) and the accumulated image data stored in the image data holding means (24).
An image data restoring device characterized by having c). 5. The image data restoring device according to claim 3, wherein said accumulation operation means subtracts a value obtained by raising 2 to the power of (N-1) from the accumulated image data from the image data holding means. And a switching circuit (32e) for outputting the accumulated image data from the image data holding means (24) in the first hierarchy and outputting the image data from the subtraction means (32d) in the second and subsequent hierarchies. And an adding means (32f) for adding the accumulated image data from the switching circuit (32e) and the divided image data from the level shift means (30).