JPS62193380A - System and device for encoding and decoding picture signal - Google Patents

Abstract

PURPOSE:To make it possible to reproduce edge parts exactly by separating edge parts in which the level changes abruptly in picture signals exactly and encoding them. CONSTITUTION:Inputted codes are separated to the first code (output code of an information storing type encoder 5) and the second code (output code of an information non-storing type encoder 8) in a separating section 10 accord ing to synchronizing signals. The first code after separation is decoded by an information storing type decoder 11 and quantized output q(i,j) is reproduced, and the second code after separation is decoded by an information non-storing type decoder 13, and quantized residual signals d'(i,j) are reproduced. The infor mation storing type decoder 11 decodes run length code and reproduces run of '0' level and 128 level and reproduces q(i,j). The information non-storing type decoder 13 decodes orthogonal transformation coefficient after quantizing encoded in block unit, and reproduces quantized residual signals d' (j,j) by mak ing reverse orthogonal transformation. Quantized output q(j,j) and quantized residual signals d'(i,j) are added by an adder 14 in block unit and decoded picture signals f(i,j) are formed.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image signal encoding device, a decoding device, and a method thereof for shortening the transmission time of image signals or reducing the storage capacity.

(Prior Art) There are two types of coding for image signals, particularly multivalued image signals, that have been used in the past: information-preserving coding and non-information-preserving coding. Information-preserving encoding does not include quantization in the encoding process, and it is possible to reproduce an image that is exactly the same as the original image through encoding and decoding processing. Examples of this method include dividing the image signal into multiple bit planes or level planes and applying encoding to binary signals such as run-length encoding for each plane, and DPCM encoding that does not include quantization processing. However, none of them can achieve a high compression rate. On the other hand, non-information preserving encoding refers to encoding that includes some kind of quantization processing during the encoding process, and the reproduced image contains quantization noise due to the encoding/decoding processing, resulting in deterioration of image quality. Predictive coding and orthogonal transform coding are generally used as this method. In predictive encoding, a prediction error signal, which is the difference between a predicted signal and an image signal, is quantized and encoded. In orthogonal transform encoding, an image signal is orthogonally transformed, and the transform coefficients are quantized and encoded. In this way, in non-information preserving encoding, it is possible to significantly compress the amount of information by applying quantization, but in general, quantization cuts the high frequency components of the image signal, causing the signal level to change rapidly. At edges, the signal level changes slowly and the edges become unclear. In order to solve this problem, M
A method has been proposed in which the SB is run-length coded and the remaining bits other than the MSB are predictively coded or orthogonally transformed coded.

For details, please refer to IEEE ICASS
P 85), pages 4.10.1 to 4.10.4, Charles F.
) published a paper titled “Ano 1 Hybrid Image Compression Technique (AHybrid Image Composition Technique)”
e Compression Technique)J
It is described in. This method focuses on the fact that the position of many edges corresponds to the signal change point of the MSB, and
By encoding the SB without distortion by run-length encoding, it is attempted to encode edges as accurately as possible. When decoding, the MSB is obtained by run-length decoding, the bits other than the MSB are obtained by predictive decoding or orthogonal transform decoding, and the image signal is obtained by adding the MSB and the bits other than the MSB. Distortion occurs in bits other than the MSB due to quantization, but since quantization is not performed on the MSB, no distortion occurs at all. Therefore, many edges are decoded without being obscured.

(Problems to be Solved by the Invention) However, the edge portion of the image signal does not necessarily correspond to the signal change point of the MSB. Assuming that the image signal can have 256 levels (8 bits) from 0 to 255, the image signal will have 128 levels corresponding to the MSB even if the original signal changes smoothly and does not include edges, as shown in FIG. 2(a). When the level is crossed, <MSB changes as shown in FIG. 2(b), and signals other than the MSB have sharp discontinuities as shown in FIG. 2(e). MS in such cases
The change in B does not correspond to any edges and defeats the purpose of run-length encoding the MSB to preserve edges. Furthermore, since discontinuities occurring in signals other than MSB include many high-frequency components, the compression efficiency is rather reduced.

An object of the present invention is to provide a highly efficient image signal encoding/decoding system and apparatus capable of accurately reproducing edge portions by accurately separating and encoding edge portions where the level rapidly changes in an image signal. Our goal is to provide the following.

(Means for Solving the Problems) The image signal encoding/decoding method of the present invention reads the image signal in block units on the transmitting side, determines the quantization characteristic from the maximum gradient value of the image signal in the block, and quantizing the image signal according to a quantization characteristic to generate a first signal;
A first code is generated by performing information preserving encoding on the first signal, a second signal consisting of a difference between the image signal and the first signal is generated, and the second signal is A second code is generated by performing information-preserving encoding on the second signal, and the receiving side generates the first signal by performing information-preserving decoding on the first code. The method is characterized in that the second signal is obtained by performing information non-preserving decoding on the second code, and the first signal and the second signal are added.

Further, the image signal encoding device of the present invention includes means for reading out an image signal in block units, means for determining a quantization characteristic from the maximum gradient value of the image signal in the block, and means for reading out the image signal in accordance with the determined quantization characteristic. means for generating a first signal by quantizing; means for generating a first code by performing information preserving encoding on the first signal; and the image signal and the first signal. and means for generating a second code by performing information non-preserving encoding on the second signal.

Further, the image signal decoding device of the present invention reads out the image signal in block units, determines a quantization characteristic from the maximum gradient value of the image signal in the block, quantizes the image signal according to the quantization characteristic, and converts the image signal into a first one. generating a signal, performing information-preserving encoding on the first signal to generate a first code, and generating a second signal consisting of a difference between the image signal and the first signal. input the first code and the second code from an image signal encoding device that generates a second code by performing information non-preserving encoding on the second signal; In a decoding device that obtains an image signal by decoding a code of The image forming apparatus is characterized by comprising means for obtaining a second signal by performing pattern decoding, and means for obtaining an image signal by adding the first signal and the second signal.

(Example) The present invention will be described in detail below with reference to the drawings. 1st
FIG. 1(a) is a block diagram showing an example of an encoding device that implements the image signal encoding/decoding method of the present invention, and FIG.
b) is a block diagram showing an example of a decoding device.

In the following explanation, first, two-level quantization is used to separate edge parts, run-length coding is used as information-preserving encoding of the quantized output signal, and non-information-preserving residual signal from which edges are separated. Type encoding is orthogonal transform encoding (discrete cosine transform, Hadamard transform, etc.)
However, these limitations for convenience of explanation do not impair generality as will be explained later. First, in the encoding device, an image signal is read out in blocks by a block reading unit 1, which performs orthogonal transformation. For example, an image signal of 8 bits per pixel is read out with a total of 256 pixels (16 pixels vertically and 16 pixels horizontally) as one block. This one block image signal f
(i, j) is sent to the parameter extraction unit 2, and the maximum gradient fdif for pixels within one block is calculated. Here, f(i, j) is the value of the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction. The gradient value is a value that indicates the magnitude of change between adjacent pixels, and the gradient value of the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction can be calculated using several calculation methods as follows. It will be done.

(1) max(If(i,j)−f(i−1,j)l,
If (i, j) - f (i, j - 1) l) (2) 1 turtle, j) - saturated - 1, j) l + 1 f (i, j) - f (i, j
-1) 1 The maximum gradient value is the maximum value of the gradient values in each pixel in the block calculated in this way. The maximum gradient fdif within the block calculated in this way is sent to the quantization characteristic determining section 3. The quantization characteristic determination unit 3 compares the maximum gradient value fdif with a predetermined threshold value T, and sets different U childization characteristics depending on the magnitude. If fdif>T, for example, the quantization characteristic shown in FIG. 3(a) is used. This separates the MSB as in the conventional method, and quantizes levels 0 to 127 to O, and levels 128 to 255 to 128. If fdif≦T, the quantization output level is set to only 0, as shown in FIG. 3(b), and the quantization characteristic is such that all pixel values in the block are converted to 0.

According to the quantization characteristics set in this way, the image signal within the block is quantized by the quantizer 4, and the quantization output q(
i, j) are output. By dividing the image signal into blocks in this way and determining the quantization characteristics using the maximum gradient value of the pixel values within the block, it is possible to create an image that exceeds the MSB and has a large gradient value, such as block 1 in Figure 4. The signal is quantized to two levels, but in the conventional method, the M
Gradual image signal level fluctuations exceeding SB are fd
If if<T, conversion to a two-level signal is not performed and the quantized output becomes a constant value of 0, and is not extracted as an edge portion. (T can be set to a value of 70, for example.) The quantized output q of the quantizer 4 calculated as above
(i, j) is encoded by the information preserving encoder 5. The consecutive lengths of the 0 level and 128 level of the quantized output are encoded using a run length encoder. The run count in this run-length encoding may be terminated within each block or may be continued between blocks.

The quantized output q(i,
j) The difference with the image signal f (i, j) is taken in the subtractor 7, and the residual signal d (i, j) = f (i, j) - g
(i, j) is calculated.

The output d(i, j) of the subtractor 7 is subjected to orthogonal transform encoding in the information preserving encoder 8. FIG. 5 shows a block diagram of an information non-preserving encoder 8 using orthogonal transformation.

First, an input signal is converted into orthogonal transform coefficients by an orthogonal transformer 81 that performs orthogonal transform such as discrete cosine transform or Hadamard transform. Next, a coefficient quantizer 82 truncates small coefficient components and quantizes other coefficient components. Finally, in the coefficient encoder 83, the magnitude of each coefficient is coded according to its frequency of occurrence (for example, Huffman code, etc.).
The sign is changed using . The first code created by the information-storing encoder 5 and the second code created by the non-information-storing encoder 8 are combined in a combining section 9 and output to a transmission path, storage device, or the like. During synthesis, a synchronization code for separation is added so that the first code and the second code can be separated.

Next, the decoding device will be explained according to the block diagram of FIG. 1(b).

First, the code input from the transmission line or the storage device is divided into a first code (output code of the information-storing encoder 5) and a second code (output code of the non-information-storing encoder 8) in accordance with the synchronization signal in the separating section 10. ). The first code after separation is decoded by the information preserving decoder 11, and the quantized output q(i
, j) is played. The second code after separation is decoded by the non-information preserving decoder B, and a quantized residual signal d'(i, j) is reproduced. Information storage type decoder 11
Then, the run-length code is decoded, the runs of the 0 level and the 128 level are reproduced, and q(i, j) is reproduced. In addition, the non-information preserving decoder 13 decodes the quantized orthogonal transform coefficients encoded in block units, and performs inverse orthogonal transform to generate the quantized residual signal a'Q,j. ). The quantized output q(i, j) and the quantized residual signal d'(Lj) are added block by block in an adder 14 to generate a decoded image signal f'(i, j). .

(f'(i, j) = q(i, j) + d"(i, j)) The generated block-by-block decoded image signal f'(i, j) is finally sent to the image storage device through the block writing section 15. The image is output to an image output device such as a computer or a printer.

By using the image signal encoding/decoding device described above, the image signal can be compressed/decoded by accurately separating the edge parts.
It is possible to reduce blurring of the wedge portion due to expansion and to realize compression/expansion processing with high compression efficiency. In the above description, two-level quantization was used as an example of quantization for separating edge portions, but three or more levels of quantization may also be used. For example, in the case of 3-level quantization, as shown in Figure 6, 0.128.255
is the quantization output level, and it is possible to use a quantization characteristic that quantizes to a level close to either one. When quantization is performed at n levels (three or more levels), the information preserving encoder generates a first code by run-length encoding n-1 level planes among n level planes.

Also, the natural 2 of mbit where 2m becomes n or (n+1)
It is also possible to run-length encode each bitplane after converting the n-level signal to base or alternating binary.

Furthermore, as the information preserving encoder, not only a run-length encoder but also other encoders that do not include quantization, such as predictive encoding that encodes prediction errors without quantization, can also be used. The non-information preserving encoder can also use not only orthogonal transform encoding but also predictive encoding including quantization. When predictive coding including quantization is used, unlike orthogonal transform coding, there is no concept of a block in coding, so the size of the block read by the block reading unit 1 can be set arbitrarily.

(Effect of the invention) By accurately separating the edge portion using the present invention,
It is possible to reduce the occurrence of blurring of edges due to compression/expansion of image signals, and to realize compression/expansion processing with high compression efficiency.

[Brief explanation of drawings]

FIG. 1(a) is a block diagram showing one embodiment of the encoding device of the present invention, FIG. 1(b) is a block diagram showing one embodiment of the decoding device, and FIGS. 2(a), (b) ), (c) is a diagram to explain the problems of the conventional method, Figure 3 is a diagram showing an example of two-level quantization characteristics, Figure 4 is a diagram showing quantization processing, and Figure 5 is a diagram showing information FIG. 6 is a block diagram of an orthogonal transform encoder as an example of a non-conservative encoder, and is a diagram showing an example of three-level quantization characteristics. In the figure, 1...Block reading unit, 2...Parameter extraction unit, 3...Quantization characteristic determining unit, 4...Quantizer, 5...
... Information preserving encoder, 7... Differentiator, 8... Information non-preserving encoder, 9... Combining section, 10... Separating section, 1
1... Information saving type decoder, 13... Information non-saving type decoder, 14... Adder, 1
5...Block writing unit, 81...Orthogonal transformer, 82...Coefficient quantizer, 83...Coefficient encoder, 72 Figure 3 (a) (b)'; +40 Block 1 Block 2o 5 Roaro Diagram input

Claims (3)

[Claims] (1) On the transmitting side, read out the image signal in block units,
determining a quantization characteristic from a maximum gradient value of an image signal in a block, quantizing the image signal according to the quantization characteristic to generate a first signal, and applying an information preserving code to the first signal; A first code is generated by converting the image signal, and a second code consisting of the difference between the image signal and the first signal is generated.
A second code is generated by generating a signal, performing non-information-preserving encoding on the second signal, and performing information-preserving decoding on the first code on the receiving side. determining the first signal by performing information-saving decoding on the second code, determining the second signal by performing information-non-preserving decoding on the second code, and adding the first signal and the second signal. An image signal encoding/decoding method that obtains an image signal by performing the following steps. (2) means for reading an image signal in block units; means for determining a quantization characteristic from the maximum gradient value of the image signal in the block; and a means for quantizing the image signal according to the determined quantization characteristic to obtain a first signal. means for generating a first code by performing information preserving encoding on the first signal; and means for generating a first code by performing information preserving encoding on the first signal;
means for generating a second signal consisting of a difference from the signal;
An image signal encoding device comprising means for generating a second code by performing information non-preserving encoding on the second signal. (3) Read the image signal in block units, determine a quantization characteristic from the maximum gradient value of the image signal in the block, quantize the image signal according to the quantization characteristic to generate a first signal, and A first code is generated by performing information preserving encoding on the first signal, a second signal consisting of the difference between the image signal and the first signal is generated, and the second signal is The first code and second code are input from an image signal coding device that generates a second code by performing information-non-preserving coding on the signal, and these codes are decoded to generate an image. In a decoding device for obtaining a signal, means for obtaining a first signal by performing information-preserving decoding on a first code, and performing information-non-preserving decoding on a second code. An image signal decoding device comprising: means for obtaining a second signal by adding the first signal and the second signal; and means for obtaining an image signal by adding the first signal and the second signal.