JPH05199415A - Two-dimensional information encoding and decoding device - Google Patents

Abstract

PURPOSE:To shorten the time required for viewer's recognition of a picture at the time of displaying encoded and decoded picture data and to scatter the distortion due to encoding and decoding in the picture. CONSTITUTION:A random address generator 40 generates an address 0 to M without mutual overlapping. A block extracting circuit 42 extracts the block (nXn picture elements) in the position specified by the generated address. A random address generator 44 generates the address in the same order as the random address generator 40. A block arranging and restoring circuit 46 arranges blocks in accordance with the generated address to restore the picture data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for encoding / decoding image data block by block, and more particularly to an image data encoding / decoding apparatus having improved image data encoding / decoding order.

[0002]

2. Description of the Related Art With the development of communication technology, image data communication has been utilized in various fields. Particularly in recent years, videophones and the like that transmit still image data have become widespread. Further, a digital still camera and the like, which digitally records still image data captured by a CCD or the like on a recording medium, have also been developed.

In such image data transmission, a so-called image data encoding / decoding device is used.

FIG. 3 shows an example of a conventional coding / decoding device. Here, an encoding device is shown in (A) and a decoding device is shown in (B).

In the encoding device shown in FIG.
The image data 10 configured for each is 8 of the raster scan.
The lines of data are read out together and the image data of eight lines are stored in the buffer 12. Then, the stored data is divided into every 8 pixels in the line direction, and is sequentially transmitted to each block composed of 8 × 8 pixels.

The DCT circuit 16 first performs a discrete cosine transform, which is an orthogonal transform, on the block 14.
This discrete cosine transform is a kind of orthogonal transform as described above, and is less likely to cause block distortion and is superior in terms of compression rate and image quality as compared with other Hadamard transforms, for example.

The block after the orthogonal transformation (to be exact, a matrix of DCT coefficients) is quantized by a quantization circuit 18 in a predetermined quantization step. The quantization step here is generally set so that the quantization step is small for low spatial frequencies, while it is large for high spatial frequencies. It That is, the quantization is performed with the quantization characteristic according to the human visual characteristic.

The quantized block is then sent to the run-length coding circuit 20, where run-length coding of "0" is performed. That is, a zigzag scan is sequentially performed in the diagonal direction from the upper left (DC component) of the DCT coefficient matrix which is a quantized block, and a portion in which "0" is continuous in the scan is encoded.

In the block after the discrete cosine transform, that is, in the matrix of DCT coefficients, D existing in the upper left corner
The C coefficient is encoded differently from the above-described encoding.
That is, well-known differential encoding is performed on the DC coefficient. Then, after further Huffman coding, it is sent to the decoding unit. On the other hand, in the decoding unit, the transmitted DC coefficient is Huffman-decoded, and then the differential coded data is decoded to restore the original block.

The Huffman coding circuit performs the well-known Huffman coding on the data that has been run-length coded. This Huffman coding is a kind of variable length coding.

On the other hand, in the decoding device shown in (B), the coded block is first input to the Huffman decoding circuit 24, where Huffman decoding is performed. After the decoding, the run-length decoding circuit 26 performs run-length decoding.

Further, the block subjected to the run length decoding is inversely quantized by the inverse quantization circuit 28 according to the same quantization characteristic as that in the quantization circuit 18. Then, the inversely quantized block is subjected to the inverse transform of the orthogonal transform in the IDCT circuit.

Therefore, the blocks thus decoded are sequentially stored in the buffer 32, and the image data for 8 lines are sequentially arranged to restore the restored image data 34.
Can be obtained.

In this conventional device, block 14
Are sequentially extracted in the image data 10 in the same order as the pixel raster scan. On the other hand, in the decoding unit, the image data 34 is restored by arranging the block data in the same order.

[0015]

However, in the above-mentioned conventional coding / decoding apparatus, the order of selecting blocks in the data is determined by the order of raster scanning of pixels. When displayed on the display, for example, the block on the first row is reproduced from the upper left to the right, and the blocks on the second row and the block on the third row are sequentially displayed from the top to the bottom. .. For this reason, it takes a considerable amount of time for the viewer to understand what the image looks like, and this poses a problem of a heavy mental burden. In other words, in the current still image transmission technology, it takes some time to display all the image data, but at the time of the display, the viewer is irritated.

Further, when a local distortion occurs in an image between encoding and decoding for some reason, the distortion is clearly recognized by the human eye because it is local, This was a cause of lowering the subjective evaluation of image quality.

Such local distortion occurs when an attempt is made to finish decoding within a certain code amount. In other words, if there is a block for which the code amount must be forcibly reduced, distortion will occur there, but when differential encoding is performed for all or part of the block as described above, the effect is Was also transmitted to the block, causing local and conspicuous distortion to the blocks in the vicinity.

Further, in order to end the coding within a certain coding amount, it is necessary to measure the coding amount before the actual coding and allocate the entire coding amount to each block according to the measured value. However, this has a problem in that the image has to be processed twice, the processing time is increased, and a memory for storing the data of the entire image is required.

FIG. 4A shows an example of the case where the distortion is generated. Here, the distortion generated in a specific block is propagated in the processing direction of the block, so that the distortion is visually very noticeable.

The present invention has been made in view of the above-mentioned conventional problems, and an object thereof is to disperse the distortion contained in image data and to provide the viewer with the contents of the image as quickly as possible. An object is to provide an image data encoding / decoding device that can be recognized.

Another object of the present invention is a high-speed encoding / decoding device which requires only one processing of an image when the image is encoded within a certain code amount. An object is to provide an encoding / decoding device with improved image quality.

[0022]

In order to achieve the above-mentioned object, the present invention provides a data group formed by arranging data designated by a two-dimensional index in a predetermined order.
An encoding device that sequentially encodes each block of row m columns, and a decoding device that receives and decodes this code, the encoding device for extracting the block from the data group A first pseudo-random number generation circuit for randomly generating addresses without overlapping each other;
The block is selected and encoded according to the address generated by the pseudo random number generating circuit, and the decoding device randomly generates the address of the block in the same order as that of the first pseudo random number generating circuit. A second pseudo random number generating circuit is included, and the order of the blocks decoded according to the address generated by the second pseudo random number generating circuit is reproduced.

The invention described in claim 2 is the same as claim 1
In the two-dimensional information coding / decoding apparatus according to the item (4), the coding apparatus includes code amount adjusting means, and dynamically adjusts the code amount based on a difference between the generated code amount and the code amount target. To do.

[0024]

According to the above structure, the encoding device selectively extracts the block of the address generated by the first random number generation circuit from the data group, and encodes the selected and extracted block.

On the other hand, the coding device decodes the coded block sent from the coding device and follows the address generated by the second pseudo random number generation circuit for the decoded block. Reproduce in order. Here, since the second pseudo random number generation circuit generates the addresses in the same order as the first pseudo random number generation circuit, the image is prevented from being distorted. That is, the original image can be reproduced.

According to such a configuration, even if a certain block is distorted in the encoding / decoding process and a plurality of other blocks are affected thereby, those distortions are scattered in the data group. Is possible.

Therefore, when the two-dimensional information coding / decoding apparatus according to the present invention is used as a coding / decoding apparatus for image data, the distortion generated in the reproduced image can be dispersed and such distortion can be eliminated. You can make it inconspicuous. That is, it is possible to maintain the apparent image quality.

Note that such a pseudo random number can be used by recording a previously determined random number in a ROM or the like and reading it at the time of encoding / decoding. Further, based on the theory of the linear congruential method or the like, a pseudo random number generation circuit can be configured by a multiplier and an adder, and a random number can be generated by calculation.

According to the second aspect of the invention, since the code amount adjusting means is provided, the code amount is dynamically adjusted from the difference between the code amount and the target code amount during the encoding process. It is possible.

As described above, since the distortion generated at the time of the encoding / decoding is dispersed in the data group, when the present invention is applied to the image encoding / decoding apparatus, it is the same as the above. As a result, the distortion can be dispersed in the entire image to make the distortion inconspicuous, and the subjective evaluation of the image quality can be improved.

[0031]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows an image data coding / decoding apparatus according to the first embodiment. The same components as those of the conventional example shown in FIG. 3 are designated by the same reference numerals, and the description thereof will be omitted.

In this embodiment, as shown in (A),
The encoding device includes a random address generator 40, a block extraction circuit 42 for extracting blocks according to the address generated by the random address generator 40,
Is provided.

The random address generator 40 generates a block address of each block in the image data 10, and does not generate overlapping addresses,
All block addresses 0 to M are sequentially generated. That is, a pseudo random number is generated. The generation of a specific address will be described later.
The block extraction circuit 42 sequentially extracts the block at the position specified by the address from the image data 10 based on the generated address. Then, the extracted block 14 is sent to the DCT circuit 16 as described above.

Here, the coefficients and the like used when random numbers are generated in the random address generator 40 are sent from the random address generator 40 via the block extraction circuit 42 prior to each block.

The block 14 sent from the block extraction circuit 42 is subjected to the discrete cosine transform in the DCT circuit 16 as described above, and is then quantized in the quantization circuit 18, and further, the run The length coding circuit 20 and the Huffman coding circuit 22 perform coding and then output to the decoding device.

In the present embodiment, the decoding device is provided with a random address generator 44 and a block array restoring circuit 46 for restoring the array of each block according to the address generated by the random address generator 44. Has been.

The random address generator 44 sequentially generates addresses 0 to M without overlapping each other in the same order as the random address generator 40 described above. In this embodiment, since the coefficient related to the address generation is input as described above before the input of the code of each block, the random address generator 44 fetches the coefficient and outputs the address based on the coefficient. Occur.

Then, the block restoration circuit 46 arranges each block at a position specified by the address according to the generated address. That is, after all blocks are distributed, the image data 34 is obtained with the same configuration as the image data 10.

Although not shown in the figure, the DC coefficient after the DCT transform is quantized and then differentially encoded in the present embodiment. On the other hand, decoding is performed, the original coefficient sequence is restored, and inverse quantization is performed.

FIG. 4B shows an example of image data restored by the block array restoring circuit 46 shown in FIG. 1B. As described above, when the arrangement of blocks is restored, the arrangement of each block is restored at random, so even if distortion occurs in a certain block (black part in the figure) It becomes possible to disperse in the data. Therefore, as can be understood from the comparison with the conventional example shown in (A), it is possible to avoid the extreme deterioration of the image quality due to the continuous distortion and to improve the apparent image quality.

Moreover, since the blocks in the image data are sequentially displayed at random in the viewer of the image data, it is possible to very quickly grasp what the original image is and it is required to be swift. In such a display, the image recognition time can be shortened. Of course, the time until the final display is completed is the same as before, but according to such random block display, for example, when about half the display is performed, about the entire image image is displayed. Is what you can grasp.

Next, the principle of generating pseudo-random numbers in the random address generator 40 and the random address generator 44 will be described.

Various methods such as a linear congruential method and a square sampling method have been considered as a method for generating a random number by a computer or the like. The linear congruential method is used in this embodiment.

As described above, the addresses are generated from 0 to M
It is necessary for all addresses up to once to occur without duplication and evenly.
If addresses are generated by the linear congruential method under such conditions, the following calculation formula can be used (provided that M = 2 ^p −1).

First, with A and B as constants, the sequence X (i) is generated by the following first equation. X (i + 1) = A * X (i) + B (1) (mod 2 ^p ) where p ≧ 2 (i = 0, 1, 2, ... 2 ^p -1) where X (0) is an initial value Give as.

Here, the definition of the congruential expression is m on the right side.
It is an operation of dividing by the number specified by od and assigning the remainder to the left side. Therefore, in order to randomly generate an address using the above equation (1), first, A,
B is an odd integer, secondly, k is an odd number, and thirdly,
It is sufficient to set A = 1 + k * 2 ^p . However, p> 2.

Therefore, under the above conditions, the above (1)
By executing the expression, it becomes possible to randomly generate addresses without overlapping each other.

Further, as the address generating method using the linear congruential method as described above, the following method can be used.

First, let X (i) be the address to be obtained.
The number of blocks in the vertical direction of the image is 2 ^p, and its address is defined as 0 to 2 ^p -1. Further, the number of horizontal blocks and 2 ^q, to define the address and 0 to 2 ^q -1.

In this case, with A and B as constants, A * X
(I) + B is found, and the remainder obtained by dividing it by 2 ^{p + q} is X (i
+1) sequentially takes an integer from 0 to 2 ^{p + q} −1.

According to such a calculation, it becomes possible to describe the address of the block by the vertical component and the horizontal component, and randomly generate the block address represented by the vertical component and the horizontal component.

In this case, the vertical and horizontal components of the address can use the upper p bits and lower q bits of X (i), or the upper q bits and lower p bits of X (i).
Bits can also be used. Alternatively, it is also possible to use p bits and q bits at specific bit positions.

In the present embodiment, as described above, in the block extraction circuit 42, before transmitting each block, the constants (for example, A and B) used in such a linear congruential method and the initial value X ( 0) has been sent. Therefore, the random address generator 44 can take in the constant and the initial value sent to the destination, and based on that, generate the random address generated earlier in the same order. Of course, it is also preferable that the constants and initial values be recorded in advance in a ROM or the like and provided in each random address generator. According to such a method, it is not necessary to send constants and initial values.

Next, the image data coding / decoding apparatus of the second embodiment will be described with reference to FIG.

In FIG. 2, an encoding device is shown in (A), while a decoding device is shown in (B). The same components as those in the first embodiment described above are designated by the same reference numerals and the description thereof will be omitted.

In the second embodiment, as shown in (A), a feedback code amount adjusting section 50 for adjusting the code amount is provided. The feedback code amount adjusting unit 50 includes a comparator 52, a PID controller 54, a quantization table correction circuit 56, and a code amount setting unit 58.

The code amount of each block after Huffman coding is sent to one input terminal of the comparator 52. The code amount setter 5 is connected to the other input terminal of the comparator 52.
The set value output from 8 is sent. Then, the comparator 52 takes the difference between the two and sends the difference value to the PID controller 54. The PID controller 54 is
A correction signal is output to the quantization table correction circuit 56 based on the difference value. Thereby, the quantization table correction circuit 56 performs the quantization table correction in the quantization circuit 18 based on the correction signal. This quantization table is one in which the quantization step is set according to the spatial frequency.Normally, the step is small for those with low spatial frequency, while it is for those with high spatial frequency. Are set so that the steps are large. Therefore, when modifying the quantization table,
By dividing each value in the table by the correction parameter α, it becomes possible to easily change the step of the table. That is, by modifying this table, it is possible to adjust the code amount after Huffman coding is performed.

The correction parameter α is added to each block for each block and sent to the decoding circuit.

Next, a specific operation of the feedback code amount adjusting section 50 will be described.

The total code amount to be defined is total, and the total number of blocks is b. Under such conditions, the i-th
When processing the th block, the comparator 52 compares the total code amount result of the coding result of the past i−1 th block with the target code amount (i−1) / b * total. Then, the error error is the next (2)
It is calculated according to the formula.

Error = result- (i-1) / b * total (2) Here, the target code amount is stored in the code amount setting unit 58.

The PID controller 54 gives a correction signal to the quantization table correction circuit 56 based on the error amount thus obtained. With this modification signal, the quantization table modification circuit 56 modifies the quantization table in the quantization circuit 18.

On the other hand, as shown in FIG.
In the decoding device of the embodiment, an inverse quantization table correction circuit 60 is provided. The dequantization table correction circuit 60 takes in the correction parameter α added to each block and sent out, and corrects the dequantization table in the dequantization circuit 28 based on the α. That is, the inverse quantization is executed by the table having the same characteristics as the characteristics of the quantization table in the quantization circuit 18. According to such an inverse quantization table correction circuit 60, it is possible to accurately reproduce the data before being quantized.

As described above, according to the configuration of the second embodiment, when the total code amount is defined, the code amount for each block can be adjusted according to the total code amount. it can.

In the adjustment of the code amount by such feedback, even if the code amount is unnecessarily reduced in a plurality of blocks connected to each other and the image quality is deteriorated, in this embodiment, the blocks are randomly selected. Since the block is selected, it is possible to disperse the blocks in which the image deterioration has occurred in the entire screen, and to make such deterioration apparently inconspicuous. Therefore, the apparent image quality can be effectively saved despite the reduction of the code amount.

Although the image data has been shown in the above description, the application of the present invention is not limited to a specific image sensor, a specific image pickup method or a specific image reproduction method, and is specified by a two-dimensional index. Any data group can be applied as long as it is a data group configured by arranging the data to be arranged in a predetermined order. For example, an integrated sensor in which output signals of a pressure distribution sensor in which small pressure sensors are two-dimensionally integrated are taken out in a predetermined order, or a change in tunnel current is two-dimensionally changed by slightly moving the probe. Imaging method such as STM (scanning tunneling microscope) that was measured and recorded,
Alternatively, a CT (Computer T) that reproduces a two-dimensional tomographic image by Radon transform from a plurality of projected one-dimensional data
The present invention can be applied to any method such as an image reproduction method using omography.

[0068]

As described above, according to the present invention,
In the encoding device, a first pseudo random number generation circuit is provided,
Since the second pseudo random number generation circuit is provided in the decoding device, even if distortion is continuously generated in a plurality of blocks, the distortion can be dispersed in the data group finally. Therefore, if the present invention is applied to an image encoding / decoding device, even if distortion occurs, the distortion can be made inconspicuous and the apparent image quality can be improved.

Further, by providing the code amount adjusting means, it is possible to adjust the code amount for each block without reducing the encoding processing speed. Even in this case, even if distortion occurs in a specific block, the distortion can be scattered in the reproduced data group.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of a coding / decoding device according to a first embodiment.

FIG. 2 is a block diagram showing an overall configuration of a coding / decoding device according to a second embodiment.

FIG. 3 is a block diagram showing an overall configuration of a conventional encoding / decoding device.

FIG. 4 is an explanatory diagram showing an example of a display image and distortion generated in a block.

[Explanation of symbols]

 40 random address generator 42 block extracting circuit 44 random address generator 46 block array restoring circuit 50 feedback code amount adjusting circuit 60 inverse quantization table correcting circuit

Claims (4)

[Claims] 1. A coding device for sequentially coding a block of n rows and m columns with respect to a data group formed by arranging data specified by a two-dimensional index in a predetermined order.
A decoding device for receiving and decoding this code, wherein the coding device randomly generates first pseudo-random number without duplicating an address for extracting the block from the data group. A circuit, the block is selected and encoded according to an address generated by the first pseudo-random number generation circuit, and the decoding device includes the block in the same order as the first pseudo-random number generation circuit. A second pseudo random number generating circuit for randomly generating the address of the block, and reproducing the order of the blocks decoded according to the address generated by the second pseudo random number generating circuit. Information encoding / decoding device. 2. The two-dimensional information encoding / decoding apparatus according to claim 1, wherein the encoding apparatus has code amount adjusting means, and determines the code amount from the difference between the generated code amount and the code amount target. A two-dimensional information encoding / decoding device characterized by being dynamically adjusted. 3. The two-dimensional information coding / decoding apparatus according to claim 2, wherein the coding apparatus includes a circuit for orthogonally transforming the block, and a quantizing circuit for quantizing the data after the orthogonal transformation. A circuit for modifying the quantization step of the quantization circuit, wherein the code amount is adjusted by changing the quantization step of the quantization circuit, and the change information is attached to the code data, The decoding device, an inverse quantization circuit that performs inverse quantization of data, a circuit that corrects the quantization step of the inverse quantization circuit based on the change information of the quantization step attached to the code data, And a circuit for orthogonally transforming the inversely quantized data, wherein the encoded data is decoded by the encoding device to decode the encoded data. 4. The two-dimensional information coding / decoding apparatus according to claim 2, wherein the coding apparatus is a circuit for orthogonally transforming the block, and a quantizing circuit for quantizing the data after the orthogonal transformation. And a circuit for clearing the quantized coefficient to zero according to a predetermined standard, the code amount is adjusted by changing the standard for clearing the quantized coefficient to zero, and the decoding is performed. The encoding device includes an inverse quantization circuit that inversely quantizes data, and a circuit that orthogonally inversely transforms the inversely quantized data, and decodes the encoded data created by the two-dimensional information encoding device 2 Dimensional information encoding / decoding device.