JPH06141189A - Picture signal decoding device - Google Patents

Abstract

PURPOSE:To provide the device which notices the background or the non- background as a local property of a picture to efficiently decode the picture signal subjected to encoding adapted to each area. CONSTITUTION:Encoded area signal, background area signal, and non- background area signal are received from a transmission line 7, and the encoded area signal is separated by a demultiplexer 8 and is applied to an area decoder 12. The area decoder 12 decodes the encoded area signal to obtain an area signal B(j, k). A demultiplexer 9 distributes the encoded background area signal and the encoded non-background area signal to a background decoder 10 and a non-background decoder 11 respectively based on the area signal B(j, k). Signals decoded by the background decoder 10 and the non-background decoder 11 are synthesized by a picture synthesizer 13 to obtain a decoded picture signal F(j, k).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for efficiently decoding an image signal.

[0002]

2. Description of the Related Art A typical conventional image signal coding method is predictive coding.
g) and transform coding (Transform Codin)
There is g). Regarding predictive coding, by William K. Pratt,
"Digital Image Processing (Digit
alImage Processing "(John Willie and Sons 1978) (pp. 637-657, 225 Predictic)
It is explained in detail in ve Coding. On the other hand, regarding transform coding, reference 1 pp 667 to 699, 23
It is explained in detail in Section 2 Transform Coding. Predictive coding and transform coding will be described below based on Reference 1.

FIG. 4 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. On the transmitting side, the original image signal F (j, k) and the prediction signal F _T (j, k)
Of the prediction error signal D
Get (j, k). D (j, k) is quantized by the quantum device 22 to obtain the level number D _N (j, k), which is sent to the transmission line 26. The inverse quantizer 24 dequantizes D _N (i, k) to obtain an inverse quantized prediction error signal D _Q (j, k). The adder 25 adds D _Q (j, k) and F _T (j, k) to obtain a locally decoded signal _FL (j, k). Predictor 23
Generates a prediction signal F _T (j, k) for the original image signal F (j, k) to be encoded next time based on the locally decoded signal corresponding to the already encoded original image signal. The receiving side receives D _N (j, k) from the transmission path, and the dequantizer 27 obtains D _Q (j, k) from D _N (j, k). The predictor 29 and the adder 28 perform exactly the same operation as on the transmitting side,
The decoded image signal F (j, k) is obtained by the final enemy. D
If there is no transmission error of _N (j, k), F on the transmitting side
_L (j, k) and F (j, k) on the receiving side are exactly the same signal.

In predictive coding, the original image signal F
(J, k) is a prediction error signal D (j,
information is reduced by converting the prediction error signal into a signal D _N (j, k) having a smaller number of levels by performing quantization on the prediction error signal.

FIG. 6 shows a basic block diagram of transform coding. The original image signal F (j, k) is divided into blocks of N × N pixels, and the converter 30

[Equation 1] By the linear transformation of N × N transform coefficients F (u, v)
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) is a row-direction conversion matrix that performs conversion along the horizontal direction of the original image signal. The coefficient selector 31 selects only part of the transform coefficient f (u, v) to obtain the selected transform coefficient f _T (u, v). The value of the coefficient not selected shall be zero.

The quantizer 32 quantizes f _T (u, v) to obtain a transform coefficient quantization level number f _TN (u, v) and sends it to the transmission line 33.

On the receiving side, an inverse quantizer dequantizes f _TN (u, v) obtained from the transmission line to obtain an inverse quantized transform coefficient f (u, v). In the inverse converter 35, f
For (u, v), every block

[Equation 2] The following linear transformation is performed to obtain a decoded image signal of one block of N × N pixels. In (2), B _C (j, u) is an inverse matrix of the matrix [A _C (j, u)], and B _R (k, v) is an inverse matrix of the matrix [A _R (k, v)].

In transform coding, all coefficients f (u, v)
From some of the coefficients f _T (u, v), and by converting the selected coefficients into a signal f _TN (u, v) having a smaller number of levels than the original image signal, the transmission information We are reducing the amount.

[0009]

To increase the information compression rate in predictive coding, an appropriate predictor may be designed to reduce the amplitude of the prediction error signal D (j, k). Therefore conventionally, to predict the _{F (j, k) F L} (j, k-
1), _FL (j-1, k-1), _FL (j-1, k +)
A method using a plurality of locally decoded signals such as 1) has been proposed. (Reference 1, pp650-657). As shown in FIG. 5, if the pixel arrangement is numbered from S ₁ to S ₁₂ with S _O = F (j, k) as the center, for example, S ₁
= F _L (j, k−1), S ₂ = F _L (j−1, k), S
_{4 = F L (j-1} , k-1), S 6 = F L (j-1, k
+1). The amplitude of the prediction error signal D (j, k) can be reduced by increasing the number of locally decoded signals used to predict F (J, K) and using an appropriate prediction formula. For example,
_{S 0 = F (j, k} ) predicted signal _{S O = F T (j,} k) of the possible to create a linear combination of the S ₁ to S _{12. S O = A 0 + A 1} S 1 + A 2 S 2 + ......... + A 12 S 12 ... (3) A 0, A 1 ..., A 12 is D from statistical properties of the image signal at constant (J, k) Can be determined so that the average amplitude of Hereinafter, A _i (i = 1 to 12) is referred to as a prediction coefficient. For example, the size of A ₁ is statistically S
_It is large when the correlation between ₀ and S ₁ is high. Since S ₀ and S ₁ are pixels that are adjacent to each other in the horizontal direction, statistically, when the image correlation in the horizontal direction is high, the correlation between S ₀ and S ₁ is high and the value of A ₁ is large. Similarly, the size of A ₂ is statistically S ₀
The higher the correlation between S ₂ and S ₂ is, the larger the result. Since S ₀ and S ₂ are pixels that are adjacent to each other in the horizontal direction, statistically, when the image correlation is high in the vertical direction, the correlation between S ₀ and S ₂ is high, and the value of A ₂ is large.

In equation (3), A _i (i = 1 to 12) is often determined by the statistical properties of the entire image. However, if the image is locally observed around a certain S ₀ , the correlation is strong in the horizontal direction, and it is better to increase the coefficient A ₁ indicating the correlation between S ₀ and S ₁ , or the correlation is strong in the vertical direction. ₀ and S
_When it is better to increase the coefficient A ₂ indicating the correlation of ₂ , or when it is better to increase the coefficient A ₄ indicating the correlation between S ₀ and S _{4 in} the diagonally upper left direction, There are various cases in which there is a strong correlation in the direction of, and it is better to increase the coefficient A ₆ indicating the correlation between S ₀ and S ₆ , and the prediction coefficient A _i determined from the statistical properties of the entire image and the image It is usually different from the prediction coefficient A _i determined from the local property.

In the conventional predictive coding method, since the prediction coefficient A _i is set to be constant for the entire image in the equation (3), there is a drawback that the prediction which matches the local property of the image cannot be performed. In order to improve this point, the prediction coefficient Ai may be determined for each position (j, k) of each pixel based on the local properties of the surrounding image. However, in this case, the value of the prediction coefficient has to be sent from the transmission side to the encoded reception side for each pixel position (j, k), 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 f (u, v) whose amplitude is statistically small. Therefore, conventionally, the variance of the amplitude of f (u, v) is statistically obtained over the entire image, and the transform coefficient f with a small variance is obtained.
A method of truncating (u, v) has been proposed. (Reference 1, P673). Further, 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 f (u, v) is obtained for each class, and the variance is small within each class. Conversion factor f
A method of truncating (u, v) has also been proposed (Wein-Hsiung Chen).
Others "Adaptive Coating of Monochrome and Color Images (Adaptive C
oding of Monochrome and C
color Images) ”, (IEEE Trans
(actions on Communications)
Magazine November 1977 issue, pp1286-1292, referred to as reference 2 below). The coefficient truncation method in Reference 1 is
Do exactly the same for all blocks. However, when attention is paid to a certain block, the amplitude of the transform coefficient f (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 just a statistic obtained over the entire image, and the transformation coefficient f (u,
The magnitude of v) can often be much larger than its variance. As described above, the conventional transform coding method according to Document 1 has the drawback that the coefficient truncation is performed in the same manner for all blocks, and thus the coefficient truncation that matches the local property 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, in this case, it is necessary to encode which coefficient is discarded for each block and send it from the transmitting side to the receiving side, which has a drawback that the transmission code amount 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 code which coefficient is discarded for each block, and it suffices to code for each class, so the transmission code amount does not become so large. However, there is a drawback that it is not clear why the classification by AC energy leads to the coefficient truncation adapted to the local property of the image.

[0013]

According to the invention of the present application, a means for measuring a difference value for each block of an image, and an image is divided into a background area and a non-background area for each block based on the histogram of the difference value. Means, an encoding means for encoding the result of the area separation, an encoding means for the background area, and an encoding means for the non-background area. In an image signal decoding device that receives a coded image signal from a coding device, decodes the coded image signal and outputs a decoded image, inputs the coded image signal and outputs a coded region separation result. Decoding means, suitable decoding means for non-background areas, suitable decoding means for background areas, and decoding results for background and non-background areas And synthesize the whole Image signal decoding apparatus characterized by having a means for obtaining an image.

[0014]

In the image signal coding apparatus to which the present invention is applied, paying attention to the background or non-background, which is the local property of the image, and before coding the original image signal, the The background is separated to facilitate efficient coding in the subsequent coding. The present invention obtains a decoded image signal by decoding the encoded image signal received from the original image signal encoding device.

[0015]

FIG. 1 is a block diagram showing an image signal code decoding apparatus including an image signal decoding apparatus according to an embodiment of the present invention. In the image signal encoding device on the transmission side, the original image signal F (j, k) is input from the terminal 100 and added to the background separator 1, the background encoder 2 and the non-background encoder 3. The background separator 1 determines whether or not each pixel of the original image signal F (j, k) belongs to the background. B (j, k) = 1 in the background area and B (j, k) = in the non-background area. A region signal of 0 is generated and added to the background encoder 2, the non-background encoder 3, the region encoder 4, and the multiplexer 5. The background encoder 2 encodes the image signal F (j, k) belonging to the background area corresponding to B (j, k) = 1 according to its property and adds the result to the multiplexer 5. The non-background encoder 3 encodes an image signal belonging to the non-background area corresponding to B (j, k) = 0 according to its property, and adds the result to the multiplexer 5. The area encoder 4 encodes B (j, k) with a run length code or the like with a code amount as small as possible and adds it to the multiplexer 6. The multiplexer 5 has B (j, k) = 1
, The output of the background encoder 2 is selected, and B (j, k) = 0
, The output of the non-background encoder 3 is selected and added to the multiplexer 6. The multiplexer 6 first selects the output of the area encoder 4, and then selects the output of the multiplexer 5 and outputs it to the transmission line 7.

A specific example of the background separator 1 is shown in the block diagram of FIG. The original image signal F (j, k) is input from the terminal 100, the highest level detector 14 sets N × N pixels as one block, and the highest level MAX in the block is detected. On the other hand, the lowest level MIN in the block by the lowest level detector
To detect. The subtractor 20 calculates MAX-MIN in each block and adds it to the histogram measuring device 16. The histogram measuring device 16 obtains the MAX-MIN frequency distribution per image, and obtains a result as shown in FIG. 3, for example.
As can be seen from 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, the opposite is true, and the level fluctuation is large, and therefore MAX-M
IN is considered to be large. Therefore, the MA shown in Fig. 3
When the frequency distribution of the X-MIN is obtained, as a boundary d _B that give the valley of the frequency distribution, a region it than MAX-MIN is greater, the non-background area. Than that MAX-MI
The area where N is small is the background area. In Figure 2, a valley detector 17 receives the frequency distribution from the histogram measuring instrument 16 obtains the value d _B of MAX-MIN giving valley of the frequency distribution, is added to the background detector 18. BACKGROUND detector 18, the level difference MAX-MIN of each block to determine whether the difference is less than d _B, a small time B (j, k) = 1 or background, large when B (j, k) = 0 That is, it is a non-background. In the case of the background separator of FIG. 2, the background and the non-background are separated for each block of N × N pixels, so that B (j, k) has a constant value for each block. The control circuit 19 supplies a synchronization signal and a clock signal to each unit and controls them.

On the other hand, in Japanese Laid-Open Patent Publication No. 60-218991, the difference is calculated for each pixel and the region is divided as accurately as possible based on the histogram. It becomes smooth as shown in the figure. Further, for example, a fingerprint image as shown in FIG. 7 (256 pixels × 320 pixels for one finger, 32 dotted lines)
In the case of a boundary of a block of 32 pixels and a region surrounded by a solid line is a solid line fingerprint portion), it is preferable to encode the real fingerprint portion and the background portion separately, which is disclosed in JP-A-60-21899.
When you want to separate the actual fingerprint part and the background accurately as in 1, it is better to separate them along the outermost ridge,
The boundary line becomes a smooth line image, and the code amount when encoding it is not small at all. On the other hand, in the example shown here, since it is determined whether the actual fingerprint portion or the background portion is determined for each block, one block (32 × 32 pixels)
The boundaries can be encoded with bits. The size of one block is
For example, when block coding (two-dimensional discrete individual sine transform or the like) is used for the real fingerprint portion, coding control may be facilitated by setting the block size to be the same as or a multiple or divisor thereof.

As the background encoder 2, first, it is conceivable to obtain the average value of the image signal levels in the background area for one image and output it as the encoding result. This is a method that focuses on the fact that the image signal has a small level variation in the background region. As the background encoder 2, it is possible to use the second transform coding. At this time, again, paying attention to the fact that the level fluctuation of the image signal is small in the background region, the number of coefficients selected by the coefficient selector 31 shown in FIG. It is conceivable to increase the step size of the rectifier 32 and cut off the high-frequency component having a small amplitude. It is conceivable to apply predictive coding to the third background encoder 2. At this time, it is preferable to select the value of the prediction coefficient A _i that matches the statistical property of the background area that the level variation is small. Although various other background encoders can be considered, it is advantageous to use the property that the level fluctuation of the image signal is small in the background area.

As the non-background encoder 3, first, it is possible to use a transform code. At this time, paying attention to the fact that the level fluctuation of the image signal is larger than that of the non-background area, the coefficient selected by the coefficient selector 31 shown in FIG. 6 may include one corresponding to a higher frequency component. It is conceivable that the step size of the quantizer 32 is set to be small and the high frequency part having a small amplitude is encoded as faithfully as possible. Secondly, as the non-background encoder 3, it is possible to apply the predictive coding. At this time, it is advisable to select the value of the prediction coefficient A _i that matches the statistical property of the non-background region that the level fluctuation is large. Although various other non-background encoders may be considered, it is advantageous to use the property that the level fluctuation of the image signal is large in the non-background area.

Returning to FIG. 1, the receiving side of the image signal encoding / decoding apparatus, that is, the image signal decoding apparatus according to an embodiment of the present invention will be described. On the receiving side, from the transmission line 7,
Coded region signal, coded background region signal,
The encoded non-background region signal is received, and the region signal encoded by the day multiplexer 8 is applied to the separated region decoder 12. The area decoder 12 decodes the encoded area signal to obtain an area signal B (j, k). Of course, the area decoder 12 corresponds to the area encoder 4, such as a run-length decoder when the area decoder 4 is a run-length encoder. The day multiplexer 9 receives the encoded background area signal and the encoded non-background area signal from the day multiplexer 8, and based on the area signal B (j, k), outputs the encoded background area signal. In addition to the decoder 10, the encoded signal of the non-background area is added to the background decoder 11. The background decoder 10 performs a decoding operation corresponding to the background encoder 2,
The signal of the background area is obtained and added to the image synthesizer 13. For example, when the background encoder 2 is a transform encoder, the background decoder 10 is an inverse transform decoder, and when the background encoder 2 is a predictive encoder, the background decoder 10 is a predictive decoder. Similarly for non-background decoder 1
1 performs a decoding operation corresponding to the non-background encoder 3,
The signal of the non-background area is obtained and the image synthesizer 13 is added. The image synthesizer 13 synthesizes a signal in the background area and a signal in the non-background area based on the area signal B (j, k) to obtain a decoded image signal F (j, k), and outputs the decoded image signal F (j, k) to the terminal 101.

[0021]

According to the image signal coding apparatus to which the present invention is applied, paying attention to the background or the non-background which is a local property of an image, and by performing the coding suitable for each area,
The image signal is encoded with an extremely small code amount, and in the present invention, the encoded image signal transmitted from the image signal encoding device can be efficiently decoded corresponding to each area.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an encoding / decoding device in which an embodiment of the invention of the present application is a receiving side.

FIG. 2 is a block diagram showing a specific example of the background separator in the embodiment of FIG.

FIG. 3 is a diagram showing an example of a frequency distribution of level differences for each block.

FIG. 4 is a block diagram showing an example of an encoder and a decoder for predictive coding.

FIG. 5 is a diagram showing an example of a positional relationship between a reference pixel used for prediction in prediction coding and a predicted pixel.

FIG. 6 is a block diagram showing an example of a transform coding encoder and decoder.

FIG. 7 is a diagram showing an example of area division according to the present invention.

[Explanation of symbols]

 1 Background Separator 2 Background Encoder 3 Non-Background Encoder 4 Domain Encoder 5, 6 Multiplexer 7, 26, 33 Transmission Line 8, 9 Day Multiplexer 10 Background Decoder 11 Non-Background Decoder 12 Domain Decoder 13 Image Synthesizer 14 highest level detector 15 lowest level detector 16 histogram measurer 17 valley detector 18 background detector 19 control circuit 20,21 subtractor 22,32 quantizer 23,29 predictor 24,34 dequantizer 25, 28 adder 30 converter 31 coefficient selector 35 inverse converter

Claims (1)

[Claims] 1. A means for measuring a difference value for each block of an image, a means for separating an image into a background area and a non-background area for each block based on a histogram of the difference value, and encoding a result of area separation. Receiving a coded image signal from an image signal coding apparatus having a means, a means for performing a suitable encoding for a background area, and a means for performing a suitable encoding for a non-background area, In an image signal decoding device for decoding the coded image signal and outputting a decoded image, means for inputting the coded image signal and decoding the coded region separation result, and for the non-background region Is
It has means for decoding that matches it, means for decoding that matches the background area, and means for synthesizing the decoding results of the background area and the non-background area to obtain the entire decoded image. An image signal decoding device characterized by the above.