JPH04134988A - Encoder for picture - Google Patents

Abstract

PURPOSE:To make an encoder immune from an error transmission, and to prevent a degration due to the repetition of an encoding/decoding by providing a quantizer transmitting a maximum value and a minimum value at every block directly or indirectly and making the maximum value and the minimum value as representative quantization values. CONSTITUTION:A quantization characteristic in a quantizing and encoding means 205 is determined by a predicted value P, a maximum value MAX and a minimum value MIN, and together with an encoding output, the maximum value MAX and the minimum value MIN at every block are directly transmitted or those that are prescriptively converted are transmitted. That is, not only encoding word, but a dynamic range and the minimum value at every block are transmitted as additional information, and as for the quantization representative, the maximum value and the minimum value are made one of the quantization representative. Thus, the degration due to the repetition of the encoding and decoding is eliminated, and since the maximum value and the minimum value are reproduced without being affected by the predicted value, the signal degration by the error transmission can be minimized.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a highly efficient image encoding method that reduces the amount of information per pixel of a digital image signal.

Conventional technology DPCM is well known as a highly efficient encoding method with a simple circuit.

In DPCM, a predicted value of a pixel to be encoded is obtained using a decoded value of an encoded pixel, and the difference between the predicted value and the input is nonlinearly quantized.

Further, various studies are being conducted to improve the performance of DPCM.

Predicted value adaptive quantization has been proposed as a method to improve coding efficiency (References: Matsumoto et al., “Noise Shaping...”, TV
Academic journal, Vol. 42 [c12 (1988)).

A conventional DPCM using this predictive value adaptive quantization will be described below.

FIG. 4 shows a block configuration of a DPCM encoding device using this predictive value adaptive quantization.

In FIG. 4, 401 is an input terminal for a digital image signal (word length 8 bits), 402 is a quantization encoding means, and 403
A predictor outputs a predicted value of a pixel to be encoded using a decoded value of an encoded pixel, and 404 is an output terminal for encoding output.

A digital image signal from terminal 401 is input to quantization encoding means 402 . The quantization encoding means 402 quantizes the digital image signal by adaptively switching the quantization characteristic using the predicted value, outputs the quantization representative value as a decoded value, and outputs the quantization representative value as a decoded value, and quantizes the digital image signal by adaptively switching the quantization characteristic using the predicted value. Output the codeword. The code word is sent to terminal 404 as an encoded output.
It is output from

FIG. 5 shows the quantization characteristics of the quantization encoding means 402.

In the figure, solid lines indicate quantized representative values.

In normal DPC, areas close to the predicted value are finely quantized.
It is the same as M, but the predicted value allows the quantization representative value to be within the dynamic range of the input (0 because the word length is 8 bits).
Since the quantization characteristics are switched so that the quantization characteristics exist only in

A method to further improve the coding efficiency of DPCM using predicted value adaptive quantization has been proposed (Reference: Matsumoto et al., “Study of dynamic range estimation predicted value adaptive quantization method”, 1990 Television Society Annual Conference 27-3
).

This estimates the so-called dynamic range (DR) of the maximum and minimum values of pixels in the block from neighboring pixels that have already been encoded, for a block consisting of multiple pixels.
This is a method of determining whether DR is within this estimated range using 1 bit/block of control information.

Problems that the invention aims to solve However, there were the following issues.

In other words, due to encoding distortion, the determination of dynamic range estimation may differ between the first encoding and the second encoding even for the same block.
Unlike the above, there were problems in that encoding distortion occurred even in the second encoding, and problems related to resistance to errors.

In view of the above problems, it is an object of the present invention to eliminate the above signal deterioration and improve resistance to errors.

Means for Solving the Problems The image encoding device of the present invention receives a digital image signal as input, divides the image signal into blocks each consisting of a predetermined number of pixels, and calculates the maximum value MAX and minimum value MI of the blocks.
a maximum/minimum detection means for determining N; a quantization encoding means for obtaining an encoded output by quantizing and encoding each pixel of the image signal; and a prediction means for obtaining the predicted value P of the pixel to be encoded. An image encoding device comprising: a quantization characteristic in the quantization encoding means is determined by the predicted value P, the maximum value MAX, and the minimum value MIN; The maximum value M
AX, the minimum value MIN is transmitted directly or after a predetermined conversion.

Effects The image encoding method of the present invention has the following effects and features due to the above-described configuration.

■ Since the maximum and minimum values of each block are transmitted for each block, resistance to errors can be strengthened.

(2) Since preprocessing and quantization are performed so that the maximum and minimum values do not change even if encoding is repeated, signal deterioration due to repeated encoding and decoding can be eliminated.

Embodiment First, the principle of the encoding method in the image encoding apparatus of the present invention will be explained with reference to FIG.

In FIG. 1, 1 indicates the dynamic range of the input signal within the block. Since the input signal is quantized with 8 bits, it has a dynamic range of 0 to 255, but the dynamic range of the signal within the block is DR.
is from the minimum value MIN to the maximum value MAX in the block (
DR=MAX-MIN), smaller than this. Therefore, the quantization representative value in the quantizer only needs to be within this dynamic range, and thereby the coding efficiency can be improved over the conventional general PCM.

In the input dynamic range 1, Pl and P2 are examples of predicted values (M I N ≦ PI, P 2 ≦ MAX), and the characteristics of the quantizer at each predicted value are expressed as the quantization representative value and the encoding example 2. 3 if the predicted value is Pl, the predicted value is P
Case 2 is shown in 4. Since the difference between the predicted value and the value of the pixel to be encoded is probabilistically small in many cases, encoding distortion can be reduced by arranging the quantization representative value close to the predicted value.

Therefore, in the present invention, high performance is achieved by determining the quantization coding characteristics based on the maximum value, minimum value, and predicted value for each block.

The maximum value and minimum value determine the quantization encoding characteristics,
The maximum value and the minimum value are directly or indirectly transmitted as additional information so that they do not change even after repeated decoding, and the maximum value or the minimum value is used as the quantization representative value.

By assigning fixed code words to quantized representative values equal to the maximum value and minimum value, even if a transmission error occurs, the maximum value and minimum value can be reliably restored, so that the influence of error propagation can be minimized.

FIG. 2 shows an image encoding device 5 in an embodiment of the present invention.
and a block configuration diagram of the decoding bag W6.

In the encoding device 5, 201 is a digital image signal (
202 is an input terminal for the block (data word length 8 bits).
A block divider 20 that outputs pixel data for each block of 64 pixels (for example, 8 pixels in the horizontal direction and 8 pixels in the vertical direction);
3 is a maximum/minimum detector that outputs the maximum value MAX and minimum value MIN in the block, 204 is a delay device for timing adjustment, 205 is a quantization encoding means, and 206 is a predicted value P of a pixel to be encoded from a decoded value. 207 is an output terminal of the dynamic range of each block, 208 is an output terminal that outputs the minimum value of each block, 209 is an output terminal of the code word from the quantization encoding means, 210 is a subtracter, 2
11.212 is a normalization circuit, 213 is a ROM (read only memory) that performs adaptive quantization encoding of predicted values within the normalized dynamic range DR, 214 is a denormalization circuit, 215.217 is a subtracter, 216 .218 is a ROM that performs coefficient multiplication, 219 is a ROM216.21
A ROM 220 is an adder that performs multiplication of 8 with inverse characteristics.

In the decoding device 6, 221 is an input terminal for the dynamic range DR for each block, 222 is an input terminal for the minimum value for each block, 223 is an input terminal for a code word, 224 is a decoding inverse quantization means, and 225 is the same as 206. Predictor, 226
is a block decomposer, 227 is an output terminal for the decoded digital image signal, 228 is a normalization circuit, and 229 is a ROM
213 is a ROM which has inverse characteristics and performs decoding and dequantization within the normalized dynamic range DR; 230 is a denormalization circuit; 231 is a subtracter; 232 is a ROM 216.
Same ROM as 218, 233 same ROM as ROM219
M, 234 is an adder.

The operation of the encoding device and decoding device of this embodiment configured as described above will be described below.

In the encoding device, the pixel data of a digital image signal inputted from a terminal 201 is rearranged by a block divider 102, and pixel data is output for each block.

The pixel data for each block is determined by the maximum/minimum detector 203, which detects the maximum value MAX and minimum value MIN of the pixel data for each block.
is detected.

The quantization encoding means 205 has a maximum value MAX and a minimum value MIN.
, quantizes and encodes the pixels from the block divider adaptively to the predicted value P, and outputs the code word and decoded value (quantized representative value), but it is impossible to directly realize this with one ROM. Normalization circuit 21L 212 due to difficulty in circuit scale
, R0M213, and an inverse normalization circuit 214.

A subtracter 210 subtracts the minimum value MIN from the maximum value MAX and outputs the dynamic range DR.

Since predictive value adaptive quantization is performed only within the dynamic range DR, the normalization circuit 21 uses input pixel values and predictive values.
Normalization is performed by removing the minimum value within 1.212 and multiplying the reciprocal of the dynamic range DR by ROM216.218.

The ROM 213 adaptively quantizes and encodes predicted values within the normalized dynamic range, and outputs a code word and a normalized decoded value (quantized representative value).

The denormalization circuit 214 converts the normalized decoded value into a decoded value. First, the normalized decoded value is multiplied by the dynamic range DR in the ROM 219 to restore the dynamic range, and the minimum value is added in the adder 220 to obtain the decoded value.

The decoded value becomes the input of the predictor 206. Dynamic range DR1 minimum value M as additional information transmitted for each block
IN is output from terminals 207 and 208, and a code word obtained by lidding encoding each pixel in the block is output from terminal 209.

FIG. 3 shows the quantization characteristics of the quantization encoding means.

In the figure, solid lines indicate quantized representative values.

In the decoding device, dynamic range DR1 minimum value MI, which is additional information, is transmitted from terminals 221 and 222 for each block.
A code word for each pixel is input from the terminal 223 to the decoding and inverse quantization means 224. The decoding and inverse quantization means 224 performs inverse processing of the quantization and encoding means 205,
Since it is difficult to directly implement this with one ROM, it is configured with a normalization circuit 228, a ROM 229, and a denormalization circuit 230. The normalization circuit 228 is the normalization circuit 21
1.212, and the denormalization circuit 230 is the same as the denormalization circuit 214.

ROM229 has the opposite characteristics to ROM213. R
By inputting the predicted value and code word normalized by the normalization circuit 228 to the OM 229, a normalized decoded value is output, and is changed into a decoded value by the denormalization circuit 230.

The decoded values are input to a predictor 225 and a block decomposer 226. The block decomposer converts the pixels of each block into digital image signals in the same original arrangement as the signals at terminal 201, and outputs them from terminal 227.

As described above, according to this embodiment, not only the code word but also the dynamic range and minimum value of each block are transmitted as additional information, and the maximum value and minimum value are used as the quantization representative value. By doing so, there is no deterioration due to repeated encoding and decoding, and the maximum value and minimum value are reproduced without being influenced by predicted values, so signal deterioration due to error propagation can be reduced.

In the above embodiment, the dynamic range and the minimum value are transmitted for each block as additional information, but any two of the dynamic range, the maximum value, and the minimum value may be transmitted.

Further, although the minimum value is subtracted or added in the normalization circuit and the denormalization circuit, it is clear that the maximum value may be used instead of the minimum value.

The quantization characteristics are also limited to those shown in FIGS. 1 and 3, and are not limited to those shown in FIG. 3; various other characteristics may be considered. Some examples are shown below.

It has multiple quantization characteristics and can be switched depending on the position of the pixel, etc.

In the quantization encoding means, a quantization characteristic A whose maximum quantization representative value and minimum quantization representative value are the maximum value MAX and minimum value MIN of the block, respectively, and a quantization characteristic B different from the quantization characteristic A. A quantization characteristic that allows quantization to be performed using quantization characteristic A until each pixel equal to the maximum value and minimum value is encoded a predetermined number of times, and thereafter allows quantization to be performed using quantization characteristic B.

For blocks that still have a large dynamic range, it is possible to further improve the encoding efficiency by increasing the number of representative values in quantization, that is, by increasing the number of encoding bits.

Furthermore, by setting a predetermined number of blocks as a unit of encoding, examining the distribution of dynamic range within that encoding unit, and changing the amount of code according to the size of the dynamic range, the amount of code is kept constant by feedforward. Control is possible.

For a block with an extremely small dynamic range, O-bit encoding is also possible by setting the average value of the block to the minimum or maximum value and setting the dynamic range to 0.

Values within a predetermined range determined by the maximum and minimum values are replaced with their average values, and by using the average values as the maximum and minimum values, the influence of noise can be reduced and encoding efficiency can be improved. . However, in order to prevent the new maximum and minimum values from changing during re-encoding, it is necessary to perform processing to set values within the predetermined range to values outside of the new maximum and minimum values.

Furthermore, it is conceivable to further improve coding efficiency by assigning code words with shorter code lengths to quantization representative values other than the maximum value and the minimum value, which are closer to the predicted value.

DETAILED DESCRIPTION OF THE INVENTION According to the present invention, by providing a quantizer that directly or indirectly transmits the maximum value and minimum value of each block and uses the maximum value and minimum value as representative quantization values, It is highly resistant to error propagation and does not deteriorate due to repeated encoding and decoding, so it has great practical value.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram of the principle of the encoding method of the present invention, FIG. 2 is a block diagram showing the encoding device and decoding device in an embodiment of the present invention, and FIG. 3 is the quantization encoding in the present invention. FIG. 4 is a block diagram of a DPCM encoding device using conventional predicted value adaptive quantization, and FIG. 5 is a characteristic diagram of predicted value adaptive quantization. 201...Input terminal for digital west image signal, 2
02...Block decomposer, 203...
Maximum/minimum detector, 204... Delay device, 205...
... Quantization encoding means, 206 ... Predictor, 207 ... Output terminal of dynamic range DR, 208 ... Output terminal of minimum value MIN, 2
09... Code word output terminal, 210...
・Subtractor, 211.212... Normalization circuit, 2
13...ROM, 214...Denormalization circuit, 216.218...Division ROM, 214
...Multiplication ROM. Name of agent Patent attorney Akira Kodaka 2 idiots Figure 000~m - (P-MtN) Code 1, 9, 0, 000~llt - Code 1 and 2

Claims (6)

[Claims] (1) Maximum/minimum detection means that receives a digital image signal as input, divides the image signal into blocks each consisting of a predetermined number of pixels, and determines the maximum value MAX and minimum value MIN of the block, and each pixel of the image signal. An image encoding device comprising: a quantization encoding means for obtaining a coded output obtained by quantizing and coding the quantization code; and a prediction means for obtaining the predicted value P of the pixel to be coded. The quantization characteristics in the quantization means are the predicted value P, the maximum value MAX, and the minimum value M.
IN, and the maximum value MAX and the minimum value MIN for each block along with the encoded output.
An image encoding device characterized by transmitting an image directly or after a predetermined conversion. (2) In the quantization encoding means, the quantization representative value is arranged in the range from the minimum value MIN to the maximum value MAX, and the predicted value P
2. The image encoding apparatus according to claim 1, wherein an interval between the quantization representative values is small in a portion close to . (3) In the quantization encoding means, the maximum quantization representative value and the minimum quantization representative value are each the maximum value MAX of the block.
, the minimum value MIN. (4) The image encoding apparatus according to claim (1), wherein the quantization encoding means assigns specific code words to the maximum representative value and the minimum representative value. (5) The image encoding apparatus according to claim (1), wherein the quantization encoding means assigns a short code word to a quantization representative value close to the predicted value P. (6) In a block where the number of encoding bits m is 0, the average value of pixels in the block is the decoded value, the maximum value MAX and minimum value MIN to be transmitted are the average value, and the dynamic range DR is 0. Claim (1) characterized by
The described image encoding device.