JPH0671333B2 - Image signal encoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a method and apparatus for efficiently encoding and decoding a large number of images.

(B) Related Art Predictive coding and transform coding (Transform C) are representative of conventional image signal coding methods.
oding). Regarding predictive coding, "Digital Image Processing" by William K. Pratt.
(John Willie and Sons 1987) (hereinafter referred to as Reference 1) pp637 to 657, Section 22.5 Predictive Coding. On the other hand, regarding transform coding,
1pp667 to 699, 23.2 Transform Coding is explained in detail. Predictive coding and transform coding will be described below based on Reference 1.

FIG. 11 shows a basic block diagram of predictive coding. In the following, the position coordinates of each pixel forming the image signal are indicated by (j, k). j is a coordinate indicating a vertical position, and k is a coordinate indicating a horizontal position. The i-th original image signal to be transmitted is Fi (j, k). Fi (j, k) (j = 1 to M, k
= 1 to L) corresponds to one image. Here, M is the number of pixels in the vertical direction, and L is the number of pixels in the horizontal direction. Hereafter, i will be referred to as an image number. On the sending side, the original image signal Fi
(J, k) and predicted signal Of the prediction error signal Di (j,
get k). Quantizer 37 quantizes Di (j, k) to obtain a quantization level number D _N i (j, k), and sends it to transmission line 8.
The dequantizer 38 dequantizes D _N i (i, k) to obtain a dequantized prediction error signal D _Q i (j, k). The adder 40 is D _Q i (j,
k) and Are added to obtain a locally decoded signal F _L i (j, k). The predictor 39 is a prediction signal for the original image signal Fi (j, k) which is to be encoded next time based on the locally decoded signal for the already encoded original image signal. To occur. The receiving side receives D _N i (j, k) from the transmission path, and the dequantizer 41 obtains D _Q i (j, k) from D _N i (j, k). The predictor 43 and the adder 42 operate in exactly the same way as the transmitting side, and finally the decoded image signal To get If there is no transmission error of D _N i (j, k), F _L i on the sending side
(J, k) and the receiving side Are exactly the same signal.

In predictive coding, the original image signal Fi (j, k) is converted to a prediction error signal Di (j, k) with a smaller amplitude by prediction, and the prediction error signal is quantized to obtain a signal with a smaller number of levels. The amount of information converted to D _N i (j, k) is being reduced.

FIG. 13 shows a basic block diagram of transform coding. The original image signal Fi (j, k) is divided into blocks of N × N pixels and By this linear transformation, N × N transform coefficients Fi (u, v) are obtained for each block. In (1), A _C (j, u) is a column-direction conversion matrix that performs conversion along the vertical direction of the original image signal, and A _R (k, v) performs conversion along the horizontal direction of the original image signal. It is a transformation matrix in the row direction. The coefficient selector 30 selects only part of the conversion coefficient fi (u, v) and selects the selected conversion coefficient f _T i
(U, v) is obtained and the values of the unselected coefficients are zero.

The quantizer 31 quantizes f _T i (u, v) to obtain a transform coefficient quantization level number f _TN i (u, v) and sends it to the transmission line 8.

On the receiving side, f _TN i obtained from the transmission line by the inverse quantizer 34
Dequantize (u, v) and dequantize transform coefficient To get In the inverse transformer 35, for fi (u, v), for each block And a decoded image signal for one block of N × N pixels is obtained. In (2), B _C (j,
u) is the inverse of the matrix [A _C (j, u)] and B _R (k, v) is the matrix [A
It is the inverse matrix of _R (k, v)].

In transform coding, all coefficients fi (u, v) to some coefficients f
By selecting _T i (u, v) and quantizing the selected coefficient, a signal f _TN i having a number of levels lower than that of the original image signal is obtained.
The amount of transmitted information is reduced by converting to (u, v).

(C) Problems to be Solved by the Invention To increase the information compression rate in predictive coding, an appropriate predictor may be designed to reduce the amplitude of the prediction error signal Di (j, k). For this reason, in the past, to predict Fi (j, k) A method using a plurality of locally decoded signals has been proposed (Reference 1, pp650 to 657). As shown in FIG. 12, S ₀ = Fi (j,
If the pixel arrangement is numbered from S ₁ to S ₁₂ centering on k), for example, Is. If the number of locally decoded signals used to predict Fi (j, k) is increased and an appropriate prediction formula is used, the prediction error signal Di
The amplitude of (j, k) can be reduced. For example, S ₀ = Fi (j,
k) predicted signal It is conceivable to create with a linear combination of S ₁ to S ₁₂ .

A ₀ i, A ₁ i, ..., A ₁₂ i is a constant and D
It can be set so that the average amplitude of i (j, k) becomes small. Hereinafter, Ami (m = 1 to 12) is referred to as a prediction coefficient. For example, the magnitude of A ₁ i is statistically highly correlated between S ₀ and S ₁ . Since S ₀ and S _{1 are} pixels that are adjacent to each other in the horizontal direction, statistically, when the correlation of images in the horizontal direction is high, the correlation between S ₀ and S ₁ is high and the value of A ₁ i is large. Similarly, the magnitude of A ₂ i is statistically large when the correlation between S ₀ and S ₂ is high. S ₀ and S ₂
Is a pixel adjacent in the vertical direction, and statistically, if the correlation of the image in the vertical direction is high, the correlation between S ₀ and S ₂ is high and A ₂ i
The value of becomes large.

In the expression (3), Ami (m = 1 to 12) is often determined statistically by the properties of many images. However, if the image is observed locally around a certain S ₀ , the correlation is strong in the horizontal direction and S ₀
If it is better to increase the coefficient A ₁ i that indicates the correlation with S ₁ ,
Also, the coefficient A ₂ i, which has a strong correlation in the vertical direction and shows the correlation between S ₀ and S ₂ ,
If it is better to increase the value, or if the coefficient A ₄ i, which indicates the correlation between S ₀ and S ₄ , is stronger in the diagonally upper left direction, or if the coefficient A ₄ i is stronger in the diagonally upper right direction, There are various cases where it is better to increase the coefficient A ₃ i indicating the correlation between ₀ and S ₃ , and the prediction coefficient Ami determined from the statistical properties of the entire image and the prediction determined from the local properties of the image. It is usually different from the coefficient Ami. In the conventional predictive coding method, the prediction coefficient Ami is set to be constant for all images in the expression (3). That is, it is fixed regardless of the image number i. Therefore, there is an unpredictable defect that matches the properties of each image. In order to improve this point, the prediction coefficient Ami may be determined for each position (j, k) of each pixel based on the local property of the image around it. However, in this case, the value of the prediction coefficient has to be encoded from the transmitting side for each pixel position (j, k) and sent to the receiving side, which has a drawback that the transmission code amount increases.

On the other hand, in order to increase the information compression rate in transform coding, it is sufficient to design an appropriate coefficient selector and discard the transform coefficient fi (u, v) having a small amplitude statistically. For this reason, conventionally, a method has been proposed in which the variance of the amplitude of fi (u, v) is statistically obtained over a large number of images, and the transform coefficient fi (u, v) with a small variance is truncated (Reference 1, P673). ). In order to further improve the coding efficiency, each block is divided into four classes according to the magnitude of the AC energy, and then the variance of the amplitude of the transform coefficient Fi (u, v) is calculated for each class, and the transform with a small variance within each class is obtained. A method of truncating the coefficient fi (u, v) has also been proposed (Wein-Hsiung Chen, et al., “Adaptive Coding of Monochrome and Colo”).
r Images, IEEE Transactions on Communications '' 19
(November 1977, pp1285-1292, hereinafter referred to as reference 2).
The coefficient truncation method of Reference 1 is performed in the same manner for all blocks. However, when attention is paid to a certain block, the amplitude of the transform coefficient fi (u, v) is not necessarily small because the variance obtained over the entire image is small, and may be large on the contrary. The variance is only a statistic obtained over the entire image, and the size of the transform coefficient fi (u, v) of a block is often much larger than the variance. As described above, the conventional transform coding method according to Document 1 has the drawback that the coefficients are truncated in the same manner for all blocks, and thus the coefficients that match the local characteristics of the image cannot be truncated. In order to improve this point, it may be determined for each block which coefficient should be truncated. However, by doing so, which coefficient was truncated was coded for each block,
It is necessary to send from the transmitting side to the receiving side, and there is a drawback that the amount of transmitted code increases. In Reference 2, each block is divided into four classes by AC energy, and the coefficient to be cut off is the same in each class. Therefore, it is not necessary to encode which coefficient is truncated for each block,
Since it is sufficient to encode each class, the transmission code amount does not become so large. However, the basic idea of classification of blocks and truncation of coefficients by AC energy in Reference 2 is based on the rate distortion theory. In order to apply the late distortion theory to image coding, the coding distortion must be defined quantitatively. In most cases, the mean squared error is defined as coding distortion in order to facilitate theoretical treatment. In Reference 2, when the rate distortion function is calculated, the mean squared error is defined as coding distortion. It is assumed that. However, it is not always the case that the mean square error is appropriate as coding distortion. The most important thing in encoding an image is to encode the image with the least amount of information and
This is to make the distortion as invisible as possible when humans see the encoded image, that is, to improve the subjective image quality. In this way, considering the human being as the center and considering improving the subjective image quality in consideration of the human visual characteristics, the mean square error is often not appropriate as a measure for defining the coding distortion. Therefore, the basic question about the method of Document 2 is why it is a coding method suitable for an image that is divided into classes by AC energy and the truncation of coefficients is performed so that the mean square error becomes small. As described above, the method of Reference 2 has an average of 2 even though it is complicated.
Since it is based on the assumption that the multiplication error is reduced, it cannot be said that the human visual characteristics are sufficiently taken into consideration, and the subjective image quality is insufficient.

(D) Means for Solving the Problems According to the present invention, in an image signal encoding method of predictively encoding an image signal and decoding the encoded signal, the encoding side is configured to Divide into multiple blocks, measure the histogram of the difference between the highest level and the lowest level in each block for one image, analyze the shape of the histogram to determine the contrast of the input image, and input by this contrast The image is classified into a plurality of types, the classification result is coded, the quantization step size and the prediction coefficient that are predetermined for each type are selected, and the image is coded. On the decoding side, the coded classification result is obtained. Image signal for decoding the decoded image signal belonging to each type by using the predetermined quantization step size and the prediction coefficient. Coding method is obtained.

Further, according to the present invention, in an image signal encoding method for transform-encoding an image signal and decoding the encoded signal, the encoding side divides an input image into a plurality of blocks, and The histogram of the difference between the highest level and the lowest level in is measured for one image, the contrast of the input image is determined by analyzing the shape of the histogram, and the input image is classified into multiple types by this contrast. The classification result is coded, the quantization step size and the parameter related to the transform coefficient selection which are predetermined for each type are selected, and the image is coded.The decoding side decodes the coded classification result and Decode the decoded image signal to which it belongs using the predetermined quantization step size and the parameters related to the transform coefficient selection. A method of encoding an image signal for performing is obtained.

(E) Action In the present invention, the input image is analyzed in advance before encoding to obtain the contrast of the image, the image is classified into several types according to the contrast, and then the highly efficient code suitable for each type is obtained. To convert.

(F) Embodiment FIG. 1 shows an example of a block diagram of an image signal coding / decoding apparatus according to the present invention. On the transmission side, the i-th original image signal Fi (j, k) is input from the terminal 100, and is input to the contrast detector 1, the high contrast image encoder 2, the medium contrast image encoder 3, and the low contrast image encoder 4. Add. In the contrast detector 1, the contrast Ci is determined for the i-th input original image signal Fi (j, k), and the high-contrast image encoder 2, the medium-contrast image encoder 3, the low-contrast image encoder 4, the image Add to contrast encoder 5 and multiplexer 6. The value of Ci is 0 when the image has high contrast, 1 when the image has medium contrast, and 2 when the image has low contrast. Image contrast encoder 5
C i and the region signal Bi (j, k) are unequal-length coded by Huffman coding or run-length coding and added to the multiplexer 7. The area signal Bi (j, k) will be described in detail later. The high-contrast image encoder 2 has Ci of 0.
At that time, the high contrast image is coded for the i-th input Fi (j, k), and the result is added to the multiplexer 6. The medium-contrast encoder 3 encodes the medium-contrast image when Ci is 1 to the i-th input Fi (j, k), and outputs the result to the multiplexer 6
Add to. When the Ci is 2, the low-contrast image encoder 4 encodes the low-contrast image to the i-th input Fi (j, k) and outputs the result to the multiplexer 6
Add to. The multiplexer 6 selects the output of the high-contrast image encoder 2 when Ci = 0, selects the output of the medium-contrast image encoder 3 when Ci = 1, and selects the low-contrast image encoder of Ci = 2. 4 outputs are selected and added to the multiplexer 7. The multiplexer 7 first selects the output of the image contrast encoder 5, then selects the output of the multiplexer 6 and outputs it to the transmission line 8.

The receiving side receives the coded image contrast signal, the coded region signal, and the coded image signal from the transmission line 8, and the demultiplexer 9 receives the coded image contrast signal and the region signal. The encoded image signal is separated, the encoded image contrast signal and the encoded area signal are image contrast decoder 14, and the encoded image signal is a demultiplexer.
Added to 10. The image contrast decoder corresponds to the image contrast encoder 5, and obtains an image contrast signal Ci and a region signal Bi (j, k) for the i-th image by decoding the multiplexer 15 and the high contrast image decoder. 11. Add to medium-contrast image decoder 12 and low-contrast image decoder 13. The high-contrast image decoder 11 performs a decoding operation corresponding to the high-contrast image encoder 2 when Ci = 0 to obtain a decoded image signal, and adds the decoded image signal to the multiplexer 15. Medium contrast image decoder 12
Is a multiplexer for performing a decoding operation corresponding to the medium-contrast image encoder 12 when Ci = 1 to obtain a decoded image signal.
Add to 15. When Ci = 2, the low-contrast image decoder 13 performs a decoding operation corresponding to the low-contrast image encoder 13 to obtain a decoded image signal, and applies it to the multiplexer 15. The multiplexer 15 selects the output to the high-contrast image decoder 11 when Ci = 0, the output of the medium-contrast image decoder 12 when Ci = 1, and the output of the low-contrast image decoder 13 when Ci = 2. Then, the i-th decoded image signal Fi (i, k) is output to the output terminal 101.

FIG. 2 shows another example of the image signal coding / decoding apparatus according to the present invention. On the transmitting side, the i-th original image signal Fi (j, k) is input from the terminal 100 and added to the contrast detector 1 and the image encoder 16. In the contrast detector 1, the contrast Ci for the i-th input image signal Fi (j, k)
Is added to the image contrast encoder 4. The Ci value is 0 when the image has high contrast, 1 when the image has medium contrast, and 2 when the image has low contrast. The image contrast encoder 5 encodes Ci and the region signal Bi (j, k) to unequal length by Huffman encoding or run-length encoding and adds them to the multiplexer 18. The area signal Bi (j, k) will be described in detail later. The image coding parameter setter 17 determines the quantization step size, the parameter indicating the coefficient truncation reference, and the coding parameters such as the prediction coefficient according to the values of Ci and Bi (j, k), and adds them to the image encoder 16. . The image encoder 16 encodes the i-th image signal Fi (j, k) according to the determined encoding parameter and adds it to the multiplexer 18. The multiplexer 18 first selects the output of the image contrast encoder 5, then selects the output of the image encoder 16, and outputs it to the transmission line 8.

On the reception side, the transmission line 8 receives the encoded image contrast signal and the region signal and the encoded image signal, and the demultiplexer 19 encodes the encoded image contrast signal and the region signal. The encoded image signal is separated, the encoded image contrast signal and the encoded area signal Bi (j, k) are applied to the image contrast decoder 14, and the encoded image signal is applied to the image decoder 20. The image contrast decoder 14 corresponds to the image contrast encoder 5, and by decoding the image contrast signal Ci and the region signal Bi for the i-th image.
(J, k) is obtained and added to the image decoding parameter setter 21. The image decoding parameter setter 21 provides the image decoder 20 with encoding parameters such as a quantization step size, a parameter indicating a criterion for coefficient truncation, and a prediction coefficient according to the values of Ci and Bi (j, k). Defined image decoder 20 is added.
The image decoder 20 performs an operation corresponding to the image encoder 16 according to the determined encoding parameter, decodes the encoded image signal, and decodes the i-th decoded image. Is obtained and output to the output terminal 101.

Next, an example of the contrast detector 1 is shown in FIG. i
The th th original image signal Fi (j, k) is input from the terminal 100, and the highest level detector 22 sets N × N pixels as one block, and the highest level MAX in the block is detected. On the other hand, the lowest level detector 23 detects the lowest level MIN in the block. The subtractor 28 calculates MAX-MIN in each block and adds it to the histogram measuring device 24. With the histogram measuring instrument 24, MAX-M
The IN frequency distribution is obtained for each image, and the result shown in FIG. 4, for example, is obtained. As can be seen by considering an image such as a stamp or an imprinted image of a fingerprint, the level fluctuation is small in the background portion of the image, and therefore MAX-MIN is considered to be small. In the non-background part, on the contrary, the level fluctuation is large,
Therefore, MAX-MIN is considered to be large. Therefore when the frequency distribution of the MAX-MIN, such as FIG. 4 is obtained, as a boundary d _B that give the valley of the frequency distribution, it than MAX-MIN is smaller areas non-background regions, it than MAX -The area where MIN is small is the background area. In Figure 3, a valley detector 26 receives the differential frequency distribution from the histogram measuring instrument 24 obtains the value d _B of MAX-MIN give the valley of the frequency distribution, is added to the background detector 28. The background detector 28 has a level difference MAX-M of each block.
It is determined whether IN is smaller than d _B , and when it is smaller, Bi (j,
k) = 1, that is, the background, and when large, Bi (j, k) = 0, that is, the non-background. Hereinafter, Bi (j, k) will be referred to as a region signal. In the case of the background separator of FIG. 2, since the background and the non-background are separated for each block of N × N pixels, Bi (j,
k) has a constant value for each block. The control circuit 27 supplies and controls a synchronization signal and a clock signal to each unit.

The contrast measuring device 25 receives the frequency distribution from the histogram measuring device 24, pays attention to the level difference MAX-MIN having a value larger than d _B which gives a valley of the frequency distribution, and is larger than d _B1 = d _B
Set d _B2, d _B2 than made large d _B3, level difference MAX-MIN is d
_The ratio P ₁₂ between _B1 and d _B2 and the level difference MAX-MIN is d _B3
Find the ratio P ₃ that takes a larger value. Here the level difference MAX-MIN when determining the proportion considered as a whole d _B1 or more portions, determining the ratio of it. That is, the portion shown by the solid line in FIG. 5, that is, the non-background portion is considered as the whole, and the ratio to it is calculated. And P ₁₂ is less than P _B1 and P ₃ is
When P _B2 or less, Ci = 0, that is, high contrast, P
_{When 12} is P _B3 or less and P ₃ is P _B4 or less, Ci = 2, that is, low contrast, and otherwise, Ci = 1, that is, medium contrast. Here, P _B1 , P _B2 , P _B3 , and P _B4 are predetermined values.

FIG. 6 shows an example of the image encoder 16, the image encoding parameter setting device 17, the image decoding device 20, and the image decoding parameter setting device 21 shown in FIG. In FIG. 6, on the transmitting side, the converter
The 29 receives the i-th image signal Fi (j, k) from the terminal 100, divides it into blocks of N × N pixels, performs linear conversion according to the equation (1), and converts the conversion coefficient fi (u, v ) Is added to the coefficient selector 30. The coefficient selector 30 is a parameter θi indicating a coefficient selection criterion from the image coding parameter setter 17.
(J, k) is received, a coefficient is selected based on θi (j, k), and the selected coefficient f _T i (u, v) is added to the quantizer 31. Also,
Parameter k _C , which indicates the coefficient selection range based on θi (j, k),
Find l _C and add it to the unequal length encoder 32. The quantizer 31
The quantization step size Δi (j, k) is received from the image coding parameter setter 17, and the quantization step size Δi
The coefficient f _T i (u, v) received from the coefficient selection 30 is quantized at (j, k) to obtain a quantization level number f _TN i (u, v), which is added to the unequal length encoder 32. The unequal length encoder 32 uses the coefficients f _TN i (u, v) and k quantized by Huffman coding or the like.
_C and l _C are compression-encoded to obtain a compression code and added to the terminal 104. Terminal 104 is connected to multiplexer 18 in FIG. In FIG. 6, on the receiving side, the unequal length decoder 33
Performs the operation corresponding to the unequal length encoder 32, decodes the compression code, obtains the quantized coefficients f _TN i (u, v) and k _C , l _C , and applies them to the inverse quantizer 34. The inverse quantizer 34 receives the quantization step size Δi from the image decoding parameter setting unit 21.
Receives (j, k) and dequantizes the quantized coefficient Is added to the inverse converter 35. The inverse transformer 35 receives the inversely quantized coefficient, performs linear transformation according to the equation (2), and outputs the i-th decoded image signal. Is output to the terminal 101.

Fig. 7 shows that when the size of one image is 256 pixels × 320 pixels,
It is a figure which shows the mode of division at the time of dividing into the block of NxN = 16x16 pixel. FIG. 8 is a diagram showing an example of a method of selecting a coefficient obtained by linearly converting an image in a block of 16 × 16 pixels according to the equation (1) by the coefficient selector 30, and in FIG. vertical k _C following below hatched, indicating that horizontal l _C following the following coefficients are selected.
Although various methods of defining k _C and l _C can be considered, for example, for k _{C, based} on the encoding parameter θi (j, k), Can be determined to be. That is, the coefficient in the shaded area shown in FIG. 9 is first selected. Then for l _C , Can be determined. Then, of the coefficients in the shaded area in FIG. 9, the coefficients in the horizontal direction equal to or lower than l _C are selected.

FIG. 10 shows another example of the image encoder 16, the image encoding parameter setting device 17, the image decoding device 20, and the image decoding parameter setting device 21 of FIG.

The difference from the conventional image signal coding / decoding apparatus using the prediction shown in FIG. 10 and FIG. 11 is that the quantization step size Δi (j, k) and the prediction coefficient Ami (j, k) are The contrast signal Ci and the area signal Bi are not fixed, but for each image.
It is a point that can be controlled by (j, k). This control is performed by the image coding parameter setting device 17 and the image decoding parameter setting device 21.

Let Δi (j, k), θi (j, k), Ami (j, k) be Ci and Bi.
The method defined by (j, k) will be described. First, a method for determining Δi (j, k) and θi (j, k) will be described in FIG. 6, and then Δi (j, k), Ami in FIG.
A method of determining (j, k) will be described.

In FIG. 6, Δi (j, k) can be determined as shown in Table 1, for example.

In Table 1, for example, Δ ₁ <Δ ₄ <Δ ₃ <Δ ₂ (6). The idea is to reduce the quantization step size Δi (j, k) as the contrast decreases. This idea focuses on the fact that the lower the contrast, the more easily coding distortion such as block distortion is noticeable to humans. However, irrespective of the value of Ci, the contrast is low in the background region, that is, when Bi (j, k) = 1, so Δi (j, k) is set to the smallest Δ ₁ .
Ci = 0 when the non-background area is Bi (j, k) = 0
Δi (j, k) because high contrast when is, medium contrast when Ci = 1, and low contrast when Ci = 2
Was set to Δ ₂ , Δ ₃ , and Δ ₄ from large to small, respectively. By doing so, coding distortion such as block distortion is less noticeable, and an image can be coded with good quality.

In FIG. 6, θi (j, k) can be determined as shown in Table 2, for example. In Table 2, θ ₁ <θ ₂ <θ ₃ <θ ₄ (7) Is. The idea is that the lower the contrast, the more θ
This is to reduce i (j, k) and reduce the number of coefficients to be discarded. This idea focuses on the fact that the lower the contrast, the more easily coding distortion such as block distortion is noticeable to humans. However, the background area
When Bi (j, k) = 1, the contrast is extremely low and the brightness is almost constant. Therefore, θi (j, k) is set to the maximum value θ ₄ , and the background is determined by the DC component, that is, the average brightness of the block. I thought that I could represent an area.
By doing so, coding distortion such as block distortion is less noticeable, and an image can be coded with good quality.

In FIG. 10, Δi (j, k) can be determined as shown in Table 1, for example. The idea is to reduce the quantization step size Δi (j, k) as the contrast decreases. This idea focuses on the fact that the coding distortion such as granular noise is more noticeable to humans when the contrast is lower. Since the contrast is low when the background area is Bi (j, k) = 1 regardless of the value of Ci, Δi (j, k) is set to the smallest Δi. By doing so, coding distortion such as granular noise is less noticeable, and an image can be coded with good quality.

In FIG. 10, Ami (j, k) (m = 1 to 12) can be determined as shown in Table 3, Table 4, and Table 5, for example. The idea is to set the value of the prediction coefficient from a pixel distant from the pixel of interest S ₀ to be relatively large as the contrast is low, thereby increasing the predictive predictive value. When the contrast is high, if the value of the prediction coefficient from the pixel distant from the pixel of interest S ₀ is set large, the predictive predictive value may decrease.

(G) Effect of the Invention According to the present invention, it is possible to perform encoding that matches the contrast of an image, and it is possible to compress and encode an image signal with good image quality and high efficiency.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a block diagram of an encoding / decoding device of the present invention, FIG. 2 is a diagram showing another example of a block diagram of an encoding / decoding device of the present invention, and FIG. 3 is a contrast. FIG. 4 is a diagram showing an example of a detector, FIG. 4 is a diagram for explaining a method of separating a background using a frequency distribution of level differences for each block,
FIG. 5 is a diagram for explaining a method of measuring contrast using frequency distribution of level difference for each block, and FIG. 6 is an image coder, an image coding parameter setter, an image decoder, an image decoding parameter setter. Fig. 7 shows an example, and Fig. 7 shows 1
FIG. 8 is a diagram illustrating an example in which one image is divided into blocks, FIG. 8 is a diagram illustrating an example of a coefficient selection method for each block, and FIG. 9 is a diagram illustrating an example of a coefficient selection process for each block. FIG. 10 is a diagram showing another example of an image encoder, an image encoding parameter setting device, an image decoder, and an image decoding parameter setting device, and FIG. 11 is a block diagram of a conventional encoding / decoding device using prediction. The figure which shows an example. FIG. 12 is a diagram for explaining a prediction method, and FIG. 13 is a diagram showing an example of a block diagram of a conventional coding / decoding device using transformation. In the figure, 1 ... Contrast detector 2 ... High-contrast image encoder 3 ... Medium-contrast image encoder 4 ... Low-contrast image encoder 5 ... Image-contrast encoder 6,7,8,15,18 .... Multiplexer 9,10,19 …… Demultiplexer 11 …… High-contrast image decoder 12 …… Medium-contrast image decoder 13 …… Low-contrast image encoder 14 …… Image-contrast image decoder 16 …… Image encoder 17… … Image coding parameter setter 20 …… Image decoder 21 …… Image coding parameter setter 22 …… Highest level detector 23 …… Lowest level detector 24 …… Difference histogram measuring instrument 25 …… Contrast measuring instrument 26 …… Valley detector, 27 …… Control circuit 28 …… Background detector, 29 …… Transformer 30 …… Coefficient selector, 31,37 …… Quantizer 32 …… Unequal length encoder, 33 …… … Unequal length decoder 34,38 …… Inverse Coca unit, 35 ...... inverter 36 ...... subtractor, 39 ...... predictor 40 ...... adder, 41 ...... inverse quantizer 43 ...... predictor.

Claims (2)

[Claims] 1. An image signal coding method for predictively coding an image signal and decoding the coded signal, comprising: dividing an input image into a plurality of blocks on the coding side, and The histogram of the difference between the level and the minimum level is measured for one image, the contrast of the input image is determined by analyzing the shape of the histogram, and the input image is classified into multiple types by this contrast, and the classification result , And the image coding is performed by selecting a predetermined quantization step size and prediction coefficient for each type, and the decoding side decodes the coded classification result and decodes the decoded image signal belonging to each type. For the image signal encoding method, decoding is performed using the predetermined quantization step size and the prediction coefficient. 2. An image signal encoding method for transform-encoding an image signal and decoding the encoded signal, wherein an input image is divided into a plurality of blocks on the encoding side, and the maximum value in each block is obtained. The histogram of the difference between the level and the minimum level is measured for one image, the contrast of the input image is determined by analyzing the shape of the histogram, and the input image is classified into multiple types by this contrast, and the classification result Image encoding is performed, image encoding is performed by selecting a predetermined quantization step size and a parameter related to transform coefficient selection for each type, and the decoding side decodes the encoded classification result and decodes each type. An image to be decoded by using the predetermined quantization step size and the parameters related to the conversion coefficient selection for the encoded image signal. Image signal encoding method.