JPS6338898B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a code for an encoding/decoding device that temporarily stores encoded information that has multiple types of encoding characteristics and occurs irregularly in a buffer memory and then transmits it in encoded transmission of an image signal. This invention relates to a method for switching characteristics.

Variable length encoding of television (TV) signals, etc.
DPCM (Differential Pulse Code Modulation)
In such cases, encoded information such as variable-length codes that occur irregularly is temporarily stored in a buffer memory and then sent out to a transmission path in order to be transmitted at a constant transmission bit rate. In this case, the encoded information generated is controlled by switching encoding characteristics such as quantization characteristics so that the buffer memory does not overflow or underflow. Conventionally, when switching the encoding characteristic, a signal (mode signal) indicating which characteristic was switched was transmitted, but when the initial switching is performed in a long cycle, the proportion of the mode signal is smaller than other information. Although it is very rare, if the buffer memory is controlled in a short period (for example, every sampling), the amount of information in the mode signal increases and becomes a value that cannot be ignored compared to the encoded information, and the encoded information Since switching information is inserted in between, the device becomes complicated. On the other hand, a method of switching the encoding characteristics without transmitting a mode signal (“semi-variable length encoding method”) Sawada et al. “32 Mb/s for broadcast color television signals”
There is also ``Separated DPCM Coding Method'' Research and Practical Application Report Vol. 28 No. 2 (1979) P285-P286), but this is 1
Since there is a restriction that the amount of information sent in one section is constant, there are drawbacks such as poor coding efficiency.

An object of the present invention is to eliminate the above-mentioned drawbacks in a coding method that controls the occupancy rate of a buffer memory by switching coding characteristics, and to provide a code that allows switching of coding characteristics without transmitting a mode signal indicating switching of coding characteristics. The objective is to provide a method for converting

According to the present invention, after setting an initial value for accumulating the amount of encoded information (amount of encoded information) using the occupied amount of buffer memory for each section, the accumulated value of the amount of encoded information is calculated within that section. By switching the encoding characteristic according to the control characteristic using the control characteristic, the buffer memory occupancy rate can be controlled by switching the encoding characteristic without sending switching information.

The present invention will be described in detail below with reference to the drawings.

A case will be described in which the buffer memory occupancy rate in variable length coding DPCM is controlled by switching the quantization characteristics, that is, the coding characteristics of the DPCM. The sampling frequency is _S , the transmission bit rate is
If R ^bit /s and the part used for transmitting encoded information is R _B bit/s, the average transmission bit rate per sampling sample is = R/ _S , and the average amount of transmitted encoded information is _B = R _B / Indicated by _S.

When the prediction error signal of DPCM is quantized and the quantized output signal is variable-length encoded, the word length of the variable-length code, which is the encoded information of the n-th sample within one interval, that is, the amount of encoded information is W. If it is expressed in (n) bits, the n-th differential encoded information amount DW(n) obtained by subtracting the average transmitted encoded information amount _B from the encoded information amount is DW(n) = W(n) - _B ... ...(1) is shown. If no other information to be transmitted is added, the amount of differentially encoded information is the amount of change in the amount of information stored in the buffer memory (buffer occupancy). The integrated value (cumulative differential information amount) DC(n) of the amount of differentially encoded information up to the n-th sample in one section is expressed by the following equation.

DC(n)= _o 〓 ⁱ⁼ⁿ (W(i)− _B )= _o 〓 ⁱ⁼ⁿ W(i)− _B・n ……(2) where _o 〓 ⁱ⁼ⁿ W(i) is the sign The integrated value of the converted information amount (cumulative information amount) C(n) is C(n)= _o 〓 ⁱ⁼ⁿ W(i). If the initial value of the cumulative difference information amount for one section is given by DC(o), the cumulative difference information amount DC(n) in the n-th sample is given by the following equation.

DC(n)= _o 〓 ⁱ⁼ⁿ (W(n) _-B )+DC(o)= _o 〓 ⁱ⁼ⁿ W(n) _-B・n+DC(o)......(3) This cumulative difference information amount is It can be determined if the initial value DC(o) and the encoded information amount W(n) are known. That is, on the receiving side, by knowing the initial value DC(o) for each section in addition to W(n), DC(n) can be found in the same way as on the sending side. Further, if the initial value is made to match or be close to the occupancy of the buffer memory, DC(n) will become a value approximately close to the occupancy of the buffer memory. From this, it is possible to control the buffer memory occupancy rate by changing the amount of encoded information generated by switching the quantization characteristic, which is the encoding characteristic, according to the control characteristic using the value of DC(n). . As shown in Figure 1, the area of cumulative difference information is defined by constants B ₁ and B ₂ (B ₁
<B ₂ ) (generally, it can be divided using a function (control function) of n instead of a constant), the DC at the nth sample
Depending on which region (n) is in, it is determined with which encoding characteristic the next sample point (n+1) is encoded. In other words, the amount of differential information is small and DC(n)
If <B ₁ , mode 1; if B ₁ DC(n)<B ₂ , mode 2; if B ₂ DC(n), large, mode 3.

Now, the encoder has three quantization characteristics as encoding characteristics. In mode 1, the quantization step is fine, and therefore the generated information is large. Quantization characteristic 1 is for variable length encoding. When in mode 2, the quantization step is coarser and therefore less information is generated, and variable length encoding is performed with quantization characteristic 2 for variable length encoding, and in mode 3, the amount of information is more than this. Fixed bit length q (q<
Fixed length encoding is performed using quantization characteristic 3 of R _B ). In this way, the cumulative difference information amount DC
Since (n) roughly corresponds to the buffer occupancy, it is possible to control each sampling so that when the buffer occupancy is small, there is more occurrence information, and when the buffer occupancy is large, there is less occurrence information. Become. When decoding is performed on the receiving side, it is necessary to know the cumulative difference information amount DC(n) shown by equation (3), so
In addition to W(n), it is necessary to transmit information about the initial value DC(o).

The change in the buffer memory occupancy at each end of one section (block) is expressed as follows: The buffer memory occupancy in the l-th block is B(l), and the initial value is B.
O(l)o(=B(l-1)), the amount of information required to transmit the synchronization signal at the beginning of each block is d,
bit, and the amount of encoded information in the l-th interval is Wl
(i), if the number of samples in that section is denoted by N, the buffer memory occupancy B(l) at the end of the lth block is expressed by the following equation.

B(l)= _N 〓 ⁱ⁼¹ Wl(i)+α−・N+B(l−1) …(4) Using equation (3) to replace the cumulative encoded information amount with the cumulative difference information amount, we get B (l)={DC _l (N)+ _B・N−DC _l (o)}+α
-.N+B(l-1) (5) Therefore, if the initial value DC(o) is set as shown in equation (6) below, the buffer memory occupancy rate changes for each block as shown in equation (7).

DC _l (o) = _B・N+α−・N +B(l−1) ……(6) B(l)=DC _l (N)+DC _l (o) ……(7) Therefore, the accumulation within the block Differential information amount DC _l
By appropriately controlling (n), the buffer memory occupancy rate can be appropriately controlled to prevent overflow or underflow. It is not necessary that DC(n) and the buffer memory occupancy correspond to each other completely according to equation (7), and for example, the value of DC _l (o) may be a quantized value to some extent. i.e. quantized
If the value of DC _l (o) is D〓C _l (o), then D〓C _l (o) = DC _l (o) + ΔE _l (where ΔE _l is a quantization error) ...(8). Therefore, in equation (3), the quantized value D〓C _l (o) is used instead of DC(o), and from equation (9), the first
The mode is determined based on the characteristics shown in the figure, and the generated information is controlled by switching the encoding characteristics.

DC _l (n) = _o 〓 ⁱ⁼ⁿ W _l (n) − _B・n+D〓C _l (o) ...(9) In other words, since W _l (n) is obtained from the encoded information, it is cut for each section. If information about the initial value D〓C _l (o) is transmitted, switching of the encoding characteristics can be controlled for each sampling without switching information within that section. Since the initial value is sent for each section, even if there is a transmission error, the next section can be correctly reproduced, and the size of one section can be controlled without having to be fixed at a constant value of N samples. Further, the value of _B does not need to match _B = -α/N. It is also possible to perform control at appropriate sampling intervals without performing control for each sampling. The cumulative difference information amount can be calculated by accumulating the difference encoded information amount (W(n) - _B ), or from the cumulative value of the encoded information amount W(n).
It can be obtained according to equation (3).

In addition, as an example of a configuration in which control is performed using the amount of encoded information generated, a configuration is shown in which the cumulative difference information amount and the control characteristics are compared and the mode is switched, but other types of configurations are possible. (n)(=
The mode can be determined by comparing _o 〓 ⁱ⁼ⁿ W(n)) with the control characteristics.

For example, in the previous example, the area that becomes mode 1 is DC(n)
<B ₁ , but using formula (3), C(n)= _o 〓 ⁱ⁼ⁿ W(n)<B ₁ + _B・n−DC(o) ……(10−2) . Therefore, if we use the control function F ₁ (n)=B ₁ + _B・n−DC(o) instead of B ₁ as the control characteristic, then from the cumulative information amount C(n) and the control function F ₁ (n), C Mode 1 is determined by (n)<F ₁ (n). That is, in general, the cumulative amount of information and one or more control functions F
The mode can be determined by the control characteristic having (n).

A block diagram showing the configuration of the first embodiment of the present invention is shown in FIG. In this example, the sampling frequency _S =
_{A /}
Direct DPCM is performed on the D-converted NTSC color TV signal using a high-order prediction relationship, and the quantized output is variable-length coded and sent to the transmission path at 32.064 Mb/S. On the receiving side, the received signal is decoded and decoded. The case where a signal is obtained is shown below.

If one section (one block) is one horizontal scanning section, N=566 samples = 3.6 bi/s. The color TV signal A/D converted at a sampling frequency _S = 8.906 MHz is encoded in a DPCM encoder 214 in which the quantization characteristics of the quantizer 202 are switched by a mode signal from the controller 206. The prediction error signal quantized by the quantizer 202, that is, the quantized output signal, is sent to the variable length encoder.

The DPCM encoder 214 consists of a subtracter 201, a quantizer 202, an adder 203, and a predictor 204, and the quantizer 202 has quantization characteristics for variable length encoding with fine quantization steps and coarse quantization steps. It has encoding characteristics consisting of three types of quantization characteristics, two types and one type for 3.5-bit fixed length encoding, and is switched by a mode signal. As a specific example of a prediction function P(Z) that can directly and efficiently predict an NTSC color TV signal, the predictor 204 has a transfer characteristic expressed by Z transformation as expressed by the following equation. However, Z=
ej2π/s.

P(Z)=0.875Z ^-2 +0.502Z ^-3 -0.533Z ^-4 +0.156Z ^-5 ...(10) The variable length encoder 205 generates a quantized output signal according to the mode signal from the controller 206. This encoded signal (variable length encoded signal) and a code length signal indicating the word length of the code are output. The variable-length coded signal in which the word length of the code changes is sent to the buffer memory 207 and once stored, an initial value signal indicating the initial value of the amount of difference generated information for each block, a block signal indicating the head of the block, and other necessary signals are sent to the buffer memory 207 and stored therein. A signal is added and sent out onto the transmission path at a constant transmission bit rate. An occupancy signal indicating the occupancy of the buffer memory for each block is sent to the controller 206.

The controller 206 controls encoding using the code length signal sent from the variable length encoder 205 for each sample (sampled pixel) and the buffer memory occupancy signal sent from the buffer memory 207 for each block. Now the buffer memory capacity is 2048
A constant value is used as the control function of the control characteristic, and the constants are B ₁ = 768 and B ₂ = 1536.
Then, the cumulative difference information amount DC(n) determines the mode within each block according to the control characteristics shown in FIG. 3, and switches the encoding characteristics. Control for each sampled pixel is also possible, but to avoid frequent mode switching, mode switching is controlled every 32 samples, for example. If DC(n) is less than 768, mode 1; if it is 768 or more and less than 1536, mode 2;
If it is 1536 or more, it will be mode 3. For example, DC(n) is
If it becomes less than 64, a dummy code etc. is added to prevent underflow.

If the average amount of transmission coding information allocated to one sampling pixel is _B = 3.5625 (=9/16), and the information necessary for the block synchronization signal, initial value signal, etc. is α = 32 bits, then the l-th The cumulative difference information DC _l (n) in a block is expressed by equation (9) as DC _l (n) = _o 〓 ⁱ⁼ⁿ (W(i) - 3.5625) + D 〓 C (o) ... (11) . For example, when l=10, if the buffer capacity at the end of the previous block is O(l-1)=1060, then DC _l (o)=1070·775 from equation (6). maximum
If an initial value signal varying up to 2048 is transmitted in 8 bits, DC _l (o) = 1070.775 is quantized with a size of 8 and becomes D〓C _l (o) = 1064. Therefore 8
The initial value signal transmitted in bits is from 1064.

The mode signal determined by equation (11) and the control characteristics shown in FIG.
sent to. In the quantizer 202, in mode 1, quantization is performed using the variable length quantum characteristic of fine quantization steps, in mode 2, quantization is performed using the variable length quantization characteristics of coarse quantization steps, and in mode 3, quantization is performed using the variable length quantization characteristics of coarse quantization steps. is quantized using the fixed length quantization characteristic of fixed length 3.5 bits, and the variable length encoder 205 uses mode 1.
In mode 2, variable-length encoding is performed, and in mode 3, 3.5-bit fixed-length encoding (actually, 7-bit encoding is performed in total for 2 samples).

On the receiving side, signals sent from the transmission line are temporarily stored in buffer memory 208, block signals and initial value signals are sent to controller 210, and variable length encoded signals are sent to variable length decoder 209.

The variable length decoder 209 converts the variable length encoded signal into a code length signal indicating the length of the code according to the mode signal sent from the controller 210, and the variable length encoder 205 converts the variable length encoded signal into the variable length encoded signal. A quantization level signal having a level corresponding to the level of the quantized output signal is output. The code length signal is sent to the controller 210, and the quantization level signal is sent to the high-order prediction
It is sent to the DPCM decoder 215. Controller 210
Then, in the same way as the controller 206, the cumulative difference information amount is determined from the block signal, initial value signal, and code length signal.
DC(n) is determined, this value is used to determine the mode in accordance with the control characteristics in the same way as on the transmitting side, and the mode signal is sent to the variable length decoder 209. Predictor 213 and adder 2 having similar functions as predictor 204
Consists of 12. High-order predictive DPCM decoder 2
At step 15, decoding is performed from the quantized level signal, and a decoded signal of the PCM television signal is output.

As shown above, by simply sending the cutoff value for each section, the cumulative difference information DC (n ) and use this value to switch the encoding characteristic (quantization characteristic in this case) according to the control characteristic and perform encoding, thereby making it possible to control the buffer memory occupancy rate.

In this embodiment, the characteristics and values of the prediction function, the type of encoding characteristic, the type of mode in the control characteristic, the control function for dividing the mode, etc. are not limited to these.

For example, the control function of the control characteristic can divide the mode region using a function including n instead of using a constant value, and also has hysteresis in the control characteristic so that mode switching does not occur frequently. It is also possible to Even if the size of one block is not a constant value but is made variable for each block, the same control can be achieved by performing initial setting using the block signal and the initial value signal. Furthermore, it is also possible to use not one type of variable length code but a plurality of types and switch them according to the mode.

This example shows the case of variable-length coding high-order predictive DPCM, but instead of high-order predictive DPCM, for example, Hadamard transform is used to quantize the output of each sequence with multiple types of quantization characteristics. It is also possible to variable length encode the output.
In addition, as another configuration example, when using interframe coding, the prediction error and address are encoded only where the prediction error signal exceeds a certain threshold, and the prediction error and address are encoded once stored in buffer memory and then transmitted. If the code amount is used as the encoding information, the buffer occupancy rate can be controlled by switching the encoding characteristics (for example, threshold levels of a plurality of types) in accordance with the control characteristics.

Encoding information has been used to mean information directly necessary for encoding, such as the word length of a variable length code, the address, and the amount of code when encoding prediction errors. It is also possible to integrate information such as the initial value signal and synchronization signal necessary above as encoding information, and to control switching of the encoding characteristic according to the control characteristic based on the accumulated value.

As explained above, according to the present invention, the initial state is set for each section using the occupied amount of the buffer memory, and the cumulative value of the code amount of the encoded information in the section is calculated, and this value is used to set the initial state for each section. It is possible to provide an encoding system device that performs encoding by switching the encoding characteristic according to the control characteristic, and enables switching of the encoding characteristic without transmitting encoding characteristic switching information within an interval.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram showing control characteristics, FIG. 2 is a block diagram showing an embodiment of the present invention, and FIG. 3 is a diagram showing an example of controllability. 214 is a high-order predictive DPCM encoder, 215 is a high-order predictive DPCM decoder, 201 is a subtracter, 203
and 212 is an adder, 202 is a quantizer, 20
4 and 213 are predictors, 205 is a variable length encoder, 209 is a variable length decoder, 206 and 210
is a controller, and 207 and 208 are buffer memories.

Claims (1)

[Claims] 1. In an encoding/decoding device that has multiple types of encoding characteristics and that temporarily stores encoded information that occurs irregularly in a buffer memory and then transmits it, the cumulative difference information is calculated using the amount of buffer memory occupied in each section. The initial value of the amount is set, the cumulative difference information amount within the interval is sequentially determined, and the encoding characteristics are adaptively switched within the interval according to the mode determined by the value to perform encoding, and the encoded information is generated. and an encoding device that transmits the initial value for each section, and the receiving side calculates the cumulative difference information amount using the received encoded information and the initial value sent for each section, and selects the mode determined by that value. Therefore, the present invention includes a decoding device that decodes encoded information by adaptively switching encoding characteristics within an interval, and is capable of switching encoding characteristics within an interval without transmitting switching information. An encoding/decoding device characterized by: