JP2679164B2 - Encoding device and decoding device - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to encoding and decoding used for transmitting and storing image information and audio information.

Conventional technology Digital VTRs and digital audio tapes (D
Digitalization of image signals and audio signals such as AT) is progressing. Generally, during transmission and storage, an image signal is linearly quantized into 8 bits and an audio signal is linearly quantized into 16 bits. In a normal signal, the quantization error becomes almost undetectable by performing such precision quantization.

Problems to be Solved by the Invention However, the accuracy as described above may cause a problem in the case of only a very low frequency or in the case of small fluctuation. In particular, in the case of a signal that slightly changes near the quantizer threshold value, an error due to quantization is detected.

An object of the present invention is to solve such problems of the prior art.

Means for Solving the Problems According to the present invention, when transmitting or storing each sample of an image signal or an audio signal by n bits, first, each sample is n + 1.
It is obtained from a first quantizing means for quantizing with a precision of bit or more, a predicting means for predicting a quantized value to be encoded at present from a decoded value of a code word encoded in the past, and the first quantizing means. Second quantizing means for quantizing a quantized value with a quantizer that quantizes the quantized value with n + 1-bit precision in a portion substantially equal to the predicted value obtained by the predicting means, and quantizes the other portion with n-bit precision. An encoding device comprising:

Further, the present invention provides a predicting means for predicting a decoded value of a code word to be currently decoded from a past decoded value for a decoded word coded by such an encoder, and a code to be currently decoded. A decoding device comprising: a second quantizer of an encoder and a dequantizer for performing inverse quantization of an inverse characteristic using a predicted value obtained by the predictor for a word. .

Effect According to the present invention, when the quantized value to be coded and the predicted value corresponding to the quantized value are almost equal to each other (a region where the fluctuation is very small), the present invention performs the quantization with 1-bit precision higher than the normal quantization. It This makes it difficult to see the quantization error even in the part where the quantization error is easy to detect and the fluctuation is small.
become.

Further, the quantizer of the present invention is a normal linear quantizer for the part that is not equal to the predicted value, and the quantizer as a whole is close to a linear quantizer. Therefore, even if an error occurs in the transmission line, the error is not propagated by this decoder. For the same reason, a normal linear quantized signal can be decoded by the decoder of the present invention, and conversely, a signal quantized by the encoder of the present invention can be decoded by a normal linear decoder. It is also possible to do so.

Embodiment An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of an encoding apparatus of the present invention. In FIG. 1, 1 is an input part of the present invention, 2 is a first quantizer, 3 is a predictor, 4 is a second quantizer, 5 is an inverse quantizer, and 6 is an output part.

In the present embodiment, first, the signal input from the input section 1 is linearly quantized into n bits by the first quantizer 2.
The quantized value obtained by the first quantizer 2 is further quantized to n bits by the second quantizer 4 using the predicted value obtained by the predictor 3 and output to the output section 6. At the same time, the output of the second quantizer 4 is inversely quantized in the inverse quantizer 5 using the predicted value and input to the predictor 3.

FIG. 2 is a block diagram of an embodiment of the decoding device of the present invention. Reference numeral 7 in FIG. 2 indicates the input parts 3 and 5 of the present invention
Is the same predictor and inverse quantizer as in FIG. 1, and 8 is the output part.

In this embodiment, the quantized value input from the input section 7 is first dequantized by the dequantizer 5 using the predicted value obtained from the predictor 3 and output to the output section 8.

Next, the second quantizer 4 and the inverse quantizer 5 shown in FIGS. 1 and 2 will be described. FIG. 3 shows an example of the second quantizer. In this apparatus, 9 is an input part for the first quantized value, 10 is a part for inputting a predicted value, 11 is a comparator, 12 is an offset value determiner, 13 is an adder, 14 is a 1/2 multiplier, and 15 is This is the output part.
The output of the first quantizer inputted from the input 9 is compared with the predicted value inputted from the predicted value input portion 10 in the comparator 11. Then, the result obtained from the comparator 11 and the least significant bit (representing whether it is an even number or an odd number) of the predicted value are output by the offset value determiner according to Table 1. However, A in Table 1 indicates a predicted value, and B indicates a quantized value input from the input part 9. The quantized value input from the input part 9 is added with the offset value output from the offset value determiner 12 in the adder 13 and is shifted by 1 bit in the 1/2 multiplier 14 to 1
It is multiplied by / 2 and output to the output part 15. In this way n
The output of the + 1-bit first quantizer is quantized into an n-bit quantized value.

FIG. 4 shows an example of the inverse quantizer 5. In this device, 16 is an input part of a code word, 10 is an input part for inputting a prediction value, 17
Is a doubler, 18 is a comparator, 19 is an offset value determiner, 20 is an adder, and 21 is an output part. The code word input from the input 16 is shifted by 1 bit in the doubler 17 to be doubled, and is compared with the prediction value input from the prediction value input portion 10 in the comparator 18. The result obtained from the comparator 18 and the least significant bit (representing whether it is an even number or an odd number) of the predicted value are output by the offset value determining unit 19 according to Table 2.
However, in Table 2, A indicates a predicted value and B indicates the output of the doubler 17.

The output of the doubler 17 is added with the offset value output from the offset value determiner 19 in the adder 20 and output to the output portion 21. In this way, the n-bit codeword is dequantized by the n + 1-bit quantizer.

Next, FIG. 5 shows an example of the predictor. In FIG. 5, 22 is the input of the predictor input from the dequantizer, 23 is the delay device for delaying by one sample, and 24 is the output part of the predictor. With this predictor, the decoded value one sample before becomes the predicted value of the sample to be encoded as it is. A specific example of the case where one sample is coded to 8 bits using this predictor is shown below.

First, the first row of Table 3 shows a series of quantized values when quantized with 9 bits. This corresponds to the first quantization of the present invention. The second row shows the normal 8-bit quantization, which is half the 9-bit quantization. The third line is the code word of the present invention and is shown in FIGS.
It is a result derived from the apparatus of FIG. 4, FIG. 5, and FIG.

As shown in Table 3, the conventional 8-bit quantization and the present invention differ in maximum by one level and take almost the same value.
Thus, the codeword encoded according to the present invention is compatible with conventional quantization. Therefore, it is possible to decode the information coded by the present invention with a conventional device and vice versa. Similarly, according to the present invention, the past decoded value is used in the predictor, but as can be seen from the fact that the codeword is almost the same as the conventional quantization, there is almost no influence by the transmission path error. This is because the effect of the predicted value on the offset value is only two levels (with 9-bit precision) as in Tables 1 and 2.

Next, Table 4 shows the quantized values in Table 3 which are inversely quantized (decoded) to 9-bit precision. Table 5 shows the difference (quantization error) between the result obtained in Table 4 and the 9-bit precision original signal (first row).

As is clear from Table 5, the present invention reduces the quantization error as compared with the conventional 8-bit quantization. Further, the distortion of the present invention is concentrated in a place where the change is relatively large,
The quantization distortion is 0 in the small change portion. As described above, according to the present invention, it is possible to transmit a portion having a small change with 9-bit precision and a portion having a large change with 8-bit precision. Therefore, since the portion where the change in the quantizing distortion is easy to detect can be transmitted with 9-bit accuracy, the visual or auditory distortion is half that of the conventional one, which is equivalent to an S / N improvement of 6 dB. Further, the predictor of the present invention can have various configurations other than the configuration shown in FIG. 5 using a linear sum of a plurality of samples.

Next, an example in which the present invention is applied to an NTSC / TV signal will be shown. Sixth
The figure shows the configuration of the predictor of the present invention, where 22 is the input of the predictor input from the dequantizer, and 25 is one field (26
2H) Delay device to delay, 24 is the output part of the predictor.
With this predictor, the decoded value one field before becomes the predicted value of the sample to be encoded as it is. NTSC
・ For TV signals, one field (262
H) The previous sample becomes the closest in-phase point in one frame. Therefore, the predictor of FIG. 6 enables very efficient prediction.

Further, in the component TV signal, by predicting with a sample of one frame before, it is possible to transmit all still image parts with 9-bit precision and only the moving image part with 8-bit precision.

Although the present invention has been described with reference to some examples,
The device of the present invention can have various configurations other than the above embodiment. Further, a configuration in which a plurality of predictors are adaptively combined and used is also possible.

EFFECTS OF THE INVENTION According to the present invention, with the configuration as described above, when the quantized value to be encoded and the predicted value corresponding to the quantized value are close to each other (a region where the fluctuation is small), the quantization with 1-bit precision higher than that of the normal quantization is performed. Done. As a result, the quantization error is not known even in the small fluctuation portion where the quantization error is easily detected,
Visual or auditory quantization error is halved. Further, the quantizer of the present invention is a normal linear quantizer for the part that is not equal to the predicted value, and the quantizer as a whole is close to a linear quantizer. Therefore, even if an error occurs in the transmission line, the error is not propagated by this decoder. For the same reason, a normal linear quantized signal can be decoded by the decoder of the present invention, and conversely, a signal quantized by the encoder of the present invention can be decoded by a normal linear decoder. It is also possible to do so.

Further, by configuring the predictor of the present invention between frames of an image, it is possible to improve the 1-bit accuracy in the still image portion in which the quantization distortion is easily detected, as compared with normal quantization.

Finally, the device of the present invention can be realized with small hardware, and its practical effect is very large.

[Brief description of the drawings]

FIG. 1 is a block diagram of an encoding apparatus according to an embodiment of the present invention,
FIG. 2 is a block diagram of a decoding device according to an embodiment of the present invention,
FIG. 3 is a block diagram of a second quantizer of the present invention, FIG. 4 is a block diagram of an inverse quantizer of the present invention, and FIGS. 5 and 6 are block diagrams of a predictor of the present invention. 2 ... first quantizer, 3 ... predictor, 4 ... second quantizer, 5
…… Inverse quantizer, 11,18 …… Comparator, 12,19 …… Offset value determiner, 23,25 …… Delayer

Claims (4)

(57) [Claims] 1. When transmitting or storing each sample of an image signal or an audio signal in n bits, first, each sample is n + 1.
From the first quantizing means, a first quantizing means for quantizing with a precision of bit or more; a predicting means for obtaining a predicted value for a quantized value to be encoded at present from a decoded value of a code word encoded in the past; Judgment means for judging whether the obtained quantized value is within a predetermined value range with reference to the predicted value, and judging whether the quantized value is in the vicinity of the predicted value; The quantized value determined to be in is n + 1
An encoding device, comprising: a second quantizing means for quantizing with a precision of bit or more; otherwise, linear quantizing with n-bit precision. 2. The judging means has a comparing means for comparing the quantized value obtained from the first quantizing means with the predicted value obtained by the predicting means, and the second quantizing means is obtained from the comparing means. An offset value determining means for determining an offset value based on the result, an adding means for adding the offset value to the quantized value, and a 1/2 multiplying means for multiplying the output of the adding means by 1/2 are provided. The encoding device according to claim 1, characterized in that. 3. When transmitting or storing each sample of an image signal or an audio signal by n bits, first, each sample is n + 1.
First quantizing means for quantizing with a precision of bit or more; predicting means for obtaining a predictive value for a quantized value to be encoded at present from decoded values of codewords encoded in the past; Judgment means for judging whether the obtained quantized value is within a range of a predetermined value with reference to the predicted value, and judging whether the quantized value is in the vicinity of the predicted value; The quantized value determined to be in
Quantize with precision of 1 bit or more, otherwise n
A second quantization means for performing linear quantization with bit precision; and a decoded value of a codeword to be currently decoded from a past decoded value with respect to a codeword encoded by an encoding device. A second value at the time of encoding using a predictor value to be predicted and a predictive value obtained by the predictor with respect to the codeword to be currently decoded.
A decoding device, comprising: an inverse quantization means for performing quantization and inverse quantization having an inverse characteristic. 4. When transmitting or storing each sample of an image signal or an audio signal with n bits, each sample is encoded with a first quantized value quantized with an accuracy of n + 1 bits or more and previously encoded. The predicted value for the quantized value to be currently encoded is obtained from the decoded value of the codeword, and it is determined whether the first quantized value is within a predetermined value range with the predicted value as a reference. It is determined whether or not the quantized value is in the vicinity of the predicted value. If it is determined that the quantized value is in the vicinity, the first quantized value is quantized with an accuracy of n + 1 bits or more, and otherwise n An encoding method characterized by performing linear quantization with bit precision.