JPH0472434B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to an encoding apparatus using unequal length codes for image signals such as television signals. Image signals such as television signals are horizontal to the screen,
Since the correlation is strong in the vertical and temporal directions, instead of directly transmitting the image signal, a predicted signal is created from pixel signals in the vicinity of the pixel signal to be transmitted, and a difference signal between the signal to be transmitted and the predicted signal is generated. That is, by transmitting the prediction error signal, the amount of transmitted information can be saved. Such an encoding method is called predictive encoding. Since the image signals have a strong correlation, the frequency distribution of the prediction error signal is concentrated at zero. Therefore, by assigning a short code to a prediction error signal that occurs frequently, that is, a prediction error signal that has a small prediction error amplitude, and a long code to a prediction error signal that occurs less frequently, that is, a prediction error signal that has a large prediction error amplitude. For example, if the prediction error signal is encoded with unequal length, the amount of transmitted information can be saved compared to when the prediction error signal is encoded with equal length. Conventional unequal-length encoding generally uses a fixed set of unequal-length codes (for example, Huffman codes) suitable for the average frequency distribution of prediction error signals. However, in conventional methods, encoding is performed using a single type of unequal-length code suitable for the average frequency distribution. If there is a large change, the frequency distribution of the prediction error signal will also change greatly, and the frequency distribution will no longer match the unequal length code.
If the frequency distribution deviates from the average, encoding efficiency will deteriorate. On the other hand, we focused on the fact that there is still a correlation between the prediction error signals, and there are two types of codes, equal-length codes and unequal-length codes, and the correlation is determined by the size of the prediction error signal one sample before. A method of selecting which code to use and converting the code of the prediction error signal of the current sample (Hirano et al., "Equal-length/unequal-length code switching method in high-efficiency encoding of television signals", 1976 Institute of Electronics and Communication Engineers General Conference) There is a national tournament No.980). Since this method uses only the information of the previous sample, the encoding efficiency tends to deteriorate for images that have a small correlation in the horizontal direction but a strong correlation with the previous frame or in the vertical direction. SUMMARY OF THE INVENTION An object of the present invention is to provide an encoding device that can efficiently perform unequal length encoding without deteriorating encoding efficiency even if the content of an image changes significantly spatially or temporally. The encoding device of the present invention encodes a prediction error signal obtained by predictively encoding an image signal with unequal length, and converts the prediction error signal into a positive or negative signal according to a level conversion characteristic corresponding to a quantization characteristic. a level converter that converts into a quantization level number, and a plurality of encoding circuits that have assigned codes corresponding to a plurality of code types including unequal length codes and convert the quantization level numbers with the assigned codes. , For each quantization level number to be encoded next, using a plurality of spatially neighboring quantization level numbers obtained previously, it is determined for each sample which type of said assigned code is to be encoded. The present invention is characterized by comprising a code selection circuit for making a determination. According to the encoding device of the present invention, in unequal-length encoding a prediction error signal obtained by predictively encoding an image signal, the prediction error signal E obtained before the prediction error signal E to be encoded next, and which is spatially nearby The type of code to be assigned to the prediction error signal E to be encoded next is selected using a plurality of prediction error signals, and the code selection judgment is performed based on the surrounding two-dimensional or three-dimensional information including the frame direction. By making a decision with reference to the samples placed in the previous sample, it is possible to more reliably determine which type of assignment code is suitable, compared to making a decision using only the prediction error signal of the previous sample. This enables efficient encoding even if the content of the image signal changes significantly spatially or temporally. The present invention will be described in detail below using examples. FIG. 1 is a block diagram showing the configuration of a first embodiment of the present invention. In this embodiment, the unequal-length encoder has a total of three types of codes (allocation codes), one type of equal-length code and two types of unequal-length codes, and the prediction of three neighboring samples obtained previously. A case will be described in which unequal length encoding is performed while determining for each sample which type of allocation code should be selected for the prediction error signal E to be encoded next using the error signal. NTSC color TV input to input terminal 10
(TV: Television) The signal is sent to the A/D converter 5, and the sampling _{frequency s} ( _s
= _3sc ), and an 8-bit PCM signal (PCM: Pulse Code Modulation) is output and supplied to the subtracter 11 of the predictive encoder 1. The prediction signal supplied to the subtracter 11 is subtracted from the image signal supplied to the subtracter 11 by the predictor 14, and a prediction error signal is output. The prediction error signal is sent to the quantizer 12
In other words, the prediction error signal is quantized according to the quantization characteristic selected by the control signal from the control circuit 7 from among the quantization characteristics determined in advance. A quantized output level signal q is output. Note that the quantization characteristics include those in which no quantization processing is performed, that is, those in which a signal matching the input signal is output. The quantized prediction error signal is supplied to a level converter 15 and an adder 13. The adder 13 adds the quantized prediction error signal and the prediction signal supplied from the predictor 14 to output a locally decoded signal, and the locally decoded signal is sent to the predictor 14 . The predictor 14 obtains a predicted value of the next sampling time according to the prediction function P(z), and obtains a predicted signal as an output. The quantized prediction error signal sent to the level converter 15 has level conversion characteristics corresponding to the quantization characteristics selected by the control signal, for example, numbers are assigned in order from the smallest quantization level size. The quantization level numbers are converted according to the determined characteristics and are associated with positive and negative quantization level numbers. The signal converted into a quantization level number is output from the level converter 15 and is sent to the equal length encoding circuit C16, the unequal length encoding circuit A17, the unequal length encoding circuit B18, and the code selection of the unequal length encoder 2. It is sent to the decision circuit 19. The equal-length encoding circuit C16 has an equal-length code C, converts the signal of the quantization level number into a corresponding equal-length code, and supplies the same to the C terminal of the switching circuit 20. The equal length code C is suitable for a prediction error signal whose frequency distribution is a substantially uniform distribution FC. Unequal length encoding circuit A17
is an unequal length code A suitable for a prediction error signal whose frequency distribution is a distribution F _A that is largely concentrated near 0.
A circuit 2 which converts the signal of the quantization level number into the corresponding unequal length code and switches the unequal length code.
0's a terminal. Unequal length encoding circuit B18
is a distribution in which the frequency distribution of the prediction error signal is not concentrated near 0, that is, a distribution between F _A and F _C ,
It has an unequal length code B suitable for converting the quantization level number signal into a corresponding unequal length code, and sends the unequal length code to the b terminal of the switching circuit 20. The code selection decision circuit 19 determines which type of assigned code the signal with the quantization level number to be encoded next will be encoded using previously obtained signals with a plurality of spatially neighboring quantization level numbers. A code selection signal for selecting an assigned code is sent to the switching circuit 20 for each sample. In the switching circuit, one of the terminals a, b, or c is selected by the code selection signal, and the unequal length code or equal length code supplied to that terminal is sent to the buffer memory 21. Signals of equal length or unequal length codes (hereinafter referred to as transmission codes) supplied to the buffer memory 21 are used as quantization control signals, synchronization signals, etc. sent from the control circuit 7 in the buffer memory 21. Control codes necessary for decoding are added. Thereafter, it is temporarily stored in the buffer memory 21, smoothed, and output to the output terminal 22 according to the transmission speed of the transmission line.
is sent out to the transmission path. In addition, the buffer memory 21 monitors the amount of information stored in the buffer memory, and controls the amount of information stored in the buffer memory 21 by the control circuit 7.
send to In order to control the amount of generated information according to the amount of stored information sent from the buffer memory 21, the control circuit 7 performs a switching judgment of the quantization characteristic at an appropriate period, for example, every horizontal scanning period, and changes the quantization characteristic. A control signal for switching is sent to the quantizer 12, level converter 15, and buffer memory 21. The above is an explanation of the operation of the encoding device. In the predictive decoding device, information sent from the transmission path is sent from the input terminal 23 to the buffer memory 24 of the unequal length decoder 3, and the information is temporarily stored in the buffer memory 24 at the transmission speed sent from the transmission path. be remembered. The information once stored is read out sequentially according to the code length signal sent from the equal length decoding circuit C25, the unequal length decoding circuit A26, or the unequal length decoding circuit B27, and is read out based on the synchronization signal. The control signal information and the transmission code information are separated, and the quantization control signal information is sent to the level inverse converter 30, and the transmission code information is sent to the equal length decoding circuit C25 and the unequal length decoding circuit A26. The signal is supplied to the long decoding circuit B27. A code selection signal is supplied from the code selection determining circuit 29 to the equal length decoding circuit C25, the unequal length decoding circuit A26, the unequal length decoding circuit B27, and the switching circuit 28. One of the decoding circuits and the corresponding terminals a, b, or c are selected. The selected equal-length decoding circuit C25, unequal-length decoding circuit A26, or unequal-length decoding circuit B27 decodes the quantization level number corresponding to the transmission code and a code indicating the code length of the transmission code. A long signal is provided to the output of switching circuit 28. The reproduced quantization level number is supplied to the code selection determining circuit 29 and the level inverse converter 30 of the predictive decoder 4, and the code length signal is supplied to the buffer memory 24. The code selection decision circuit 29 has the same function as the code selection decision circuit 19, and performs the same processing as the code selection shaping circuit 19 using the plurality of quantization level numbers that have been unequal length decoded so far. The code selection signal at the sampling time is output. The level inverse converter 30 has a conversion characteristic opposite to that of the level converter 15, and converts the quantization level number according to the level inverse conversion characteristic selected by the control signal supplied from the control circuit 8. A quantized prediction error signal having the same quantized output level as that applied to the level converter 15 is reproduced. The reproduced prediction error signal is supplied to an adder 31 and added to the prediction signal from the predictor 32 to obtain a decoded signal. The decoded signal is sent to predictor 3
An analog decoded signal is obtained at the output of the D/A converter 6 and is supplied to the output terminal 33. The predictor 32 has the same prediction function as the predictor 14 and has the same function.
The above is an explanation of the operation of the decoding device. FIG. 2 shows a predictor 14 according to a first embodiment of the present invention.
This is a specific circuit example. The sampling frequency of the A/D converter 5 is set to three times that of the subcarrier, and a prediction function P(z) shown by equation (1) that can efficiently predict the NTSC color TV signal is used. P(z)=0.5 _z ^-1 + _z ^-3 -0.5 _z ^-4 (1) The predictor 14 has registers 40, 43, 4 that delay the input signal by one sampling clock period and output the delayed signal.
4 and 47, the adder 45, the subtracter 42, and the coefficients
This is a non-recursive type digital filter composed of multipliers 41 and 46 of 0.5. Predictor 32 is similarly configured. Next, the quantization characteristics of the quantizer 12 and the level conversion characteristics of the level converter 15 of the first embodiment of the present invention will be shown together. The quantizer 12 has two types of quantization characteristics and correspondingly two types of level conversion characteristics. Table 1 below shows the first quantization characteristic and the first level conversion characteristic only on the positive side.

[Table] The negative side is a symmetric property and is given a negative sign. Further, the combination of the second quantization characteristic and the second level conversion characteristic is shown in the following Table 2 for only the positive side. The negative side is a symmetric property and is given a negative sign.

[Table] The number of levels including positive and negative values for both the first and second characteristics is 16. On the other hand, the first and second level inversion characteristics of the level inversion converter 30 are characteristics for inversely converting each quantization level number N into its corresponding quantization output level signal q in the first and second level conversion characteristics, respectively. has. For example, the magnitude of the prediction error signal e is e=
-26, when the first quantization characteristic is selected by the control signal from the control circuit 7, the quantizer 12
The output of q = -29 quantized output level signal q,
In other words, a quantized prediction error signal is output, converted by the level converter 15 to a quantization level number N of N=-7, and supplied to the unequal length encoder 2. The quantization level number N=-7 reproduced on the receiving side is converted by the level inverse converter 30 into a quantization output level signal q of q=-29 and supplied to the adder 31. Next, the equal length code C of the equal length encoding circuit 16 and the unequal length decoding circuit A17 in the first embodiment of the present invention
A specific example of unequal length codes A and B of B18 is shown below. 16 supplied from level converter 15
The codes shown in Table 3 below are used as codes for converting the level numbers of. However, all codes shall be used to encode quantization level numbers.

[Table] Therefore, in the unequal length encoder 2, if the quantization level number N is set to N=-7 and is supplied from the level converter 15, the code selection determining circuit 1
Unequal length code A, unequal length code B, or equal length code C is selected by the code selection signal from 9, and a code of "10000001", "1000001", or "1001" is output from the switching circuit 20, respectively. . In the unequal length decoder, buffer memory 2
When the transmission signal "1001" is supplied from 4, the code selection signal from the code selection determination circuit 29
Unequal length code A, unequal length code B, or equal length code C is selected, and N=-3, N=-4, or N=-7, respectively.
The quantization level number is output from the switching circuit 28. FIG. 3 is a first concrete block diagram of the code selection determining circuit 19 in the first embodiment. In this circuit example, code selection is determined based on the quantization level numbers of a total of three samples, two samples one sample before and one almost one line before. The quantization level number signal supplied to the code selection determining circuit 19 has its absolute value taken by an absolute value circuit 48, and then is supplied to a shift register 49 with a tap.
When the sampling frequency _S is three times the subcarrier, the number of samples n _H in one horizontal scanning period is 6825. The signal input to the shift register 49 is output to the output terminal a,
The signals b and c are output after being delayed by 1,682 and 683 sampling clock cycles, respectively. The delayed and output signals are supplied to an adder 50 and added, and then supplied to a determination circuit 51. The determination circuit selects the unequal length code A if the sum of the quantization level numbers that take the absolute value is 4 or less, the equal length code C if it is 10 or more, and the unequal length code B otherwise. A code selection signal for making a determination and selecting each code is output. FIG. 4 is a second specific block diagram of the code selection determining circuit 19 in the first embodiment. This circuit example differs from the first circuit example in that the determination is made based on the combined state of three quantization level numbers instead of the determination based on the total value of the quantization level numbers. The determination circuit 52 makes a determination based on the three absolute value quantization level number signals according to a predetermined selection characteristic and outputs a code selection signal. As a selection characteristic, for example, when the size of two level numbers is 1 and the other is 2 or less, the unequal length code A is set to 2.
If the size of one level number is 5 or more, the equal length code C is selected, and in other cases, the unequal length code B is selected. FIG. 5 is a third specific block diagram of the code selection determining circuit 19 in the first embodiment. In this circuit example, code selection is determined based on a signal of a quantization level number that has been quantized. The quantization level number signal supplied to the code selection decision circuit 19 is quantized and code converted by the quantization conversion circuit 53, and is output as a converted signal having a smaller number of bits than the input signal. For example, the absolute value is 1
The level number of is converted to 0, the level number whose absolute value is 2 to 4 is converted to 1, and the level number whose absolute value is 5 or more is converted to 2. The converted signal is delayed by a shift register 49, and the delayed signals are output from terminals a, b, and c, respectively. The three output converted signals are supplied to a determination circuit 52, where a determination is made according to a predetermined selection characteristic, and a code selection signal is outputted to the output of the determination circuit 52. For example, the selection characteristic is unequal length code A when all three conversion signals are 0, and unequal length code A when the two conversion characteristics are 2.
In this case, use the note length code C, otherwise use the unequal length code B.
Let be the characteristic to be selected. Shift register 49 in FIGS. 4 and 5 has the same function as shift register 49 in FIG. As can be seen from the first to third circuit examples of the code selection circuit 19 above, the selection of the code is determined using three surrounding signals (plural signals), so compared to the case where the determination is made from one signal. A suitable code can be estimated more reliably, and therefore unequal length encoding can be performed efficiently. In terms of hardware configuration, increasing the number of samples to be referenced requires a delay circuit. It should be noted that the unequal-length encoding circuits A and B and the equal-length encoding circuit C do not need to be configured separately, and one unequal-length encoding circuit can be configured to have three types of code conversion tables, reducing the circuit size. The increase is small. As explained above, according to the encoding device of the present invention, it has multiple types of assigned codes including unequal length codes, and the prediction error signal E to be encoded next is spatially Determine for each sample which type of assignment code to select using a plurality of nearby quantization level numbers, that is, prediction error signals, and convert the prediction error signal E using the selected type of assignment code. By performing unequal length encoding, it is possible to realize an encoding apparatus that can perform encoding more efficiently than when selection is switched using only the prediction error signal of one sample before. In addition, in a device that encodes a prediction error signal obtained by predictive encoding with unequal length, if adaptive encoding is generally performed, the hardware becomes complicated, but in the present invention, By adaptively performing unequal length encoding on the prediction error signal separately from the predictive encoding part, the configuration is simple and the image content can be changed temporally with only a slight increase in hardware scale. Alternatively, it is possible to provide an encoding device that can perform encoding without deteriorating the encoding efficiency even if there is a spatial change. Predictor 14 is not limited to that shown in FIG. The code selection determining circuit is not limited to that shown in FIG. 3, FIG. 4, or FIG. 5.
Quantization characteristics of quantizer 12 and level converter 1
The level conversion characteristics of No. 5 are not limited to those shown in the first embodiment. Unequal length encoding circuit A17, unequal length encoding circuit B
18 and the codes of the equal-length encoding circuit C16 are not limited to those shown in the first embodiment. Further, in the first embodiment, the case where the types of codes are two types of unequal length codes and one type of equal length code is shown, but the present invention is not limited to this. Although the case is shown in which the switching of the quantization characteristic of the predictive encoder 1 is controlled using the information storage amount of the buffer memory in order to control the amount of generated information, the present invention is not limited to this; Control can also be performed using the cumulative amount of generated information. In the first embodiment, the case where the unequal length encoder 2 performs unequal length encoding on the signal of the quantization level number output from the predictive encoder 1 is shown, but the present invention is not limited to this. If the predictive encoder 1 is a non-recursive predictive encoder that does not have a quantizer, a prediction error signal that is not quantized is output from the predictive encoder 1, and the unequal length encoder 2 is a non-recursive predictive encoder that does not have a quantizer. The configuration is such that a prediction error signal with no difference is encoded with unequal length coding.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a first embodiment of the present invention, FIG. 2 is a block diagram showing a specific circuit example of the predictor 14, and FIG. 3 is a code selection determining circuit 1.
4 is a block diagram showing a second specific circuit example of the code selection determining circuit 19, and FIG. FIG. 2 is a block diagram showing a specific example of the circuit. 1 is a predictive encoder, 2 is an unequal length encoder, 3 is an unequal length decoder, 4 is a predictive decoder, 5 is an A/D converter,
6 is a D/A converter, 7 is a control circuit, 10 and 2
3 is an input terminal, 11 is a subtracter, 12 is a quantizer,
13, 31, 45 and 50 are adders, 14 and 32 are predictors, 15 is a level converter, 16 is equal length encoding circuit C, 17 is unequal length encoding circuit A, 1
8 is an unequal length encoding circuit B, 19 and 29 are code selection determining circuits, 20 and 28 are switching circuits, 21
and 24 are buffer memory, 22 and 3
3 is an output terminal, 30 is a level inverter, 40, 4
3, 44 and 47 are registers, 41 and 4
6 is a multiplier, 48 is an absolute value circuit, 49 is a shift register, 51 and 52 are determination circuits, and 53 is a quantization conversion circuit.

Claims (1)

[Claims] 1. A device for unequal-length encoding a prediction error signal obtained by predictively encoding an image signal, which converts the prediction error signal into positive and negative quanta according to a level conversion characteristic corresponding to a quantization characteristic. a level converter that converts the quantization level number into a quantization level number, and a plurality of encoding circuits that have assigned codes corresponding to a plurality of code types including unequal length codes and convert the codes of the quantization level numbers using the assigned codes; For each quantization level number to be encoded next, it is determined for each sample which type of said assigned code is to be encoded using a plurality of spatially neighboring quantization level numbers obtained previously. 1. An encoding device comprising: a code selection circuit.