JPH10224227A - Adaptive quantization system for orthogonal transformation coding - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain good coding with a quantization characteristic having a small dynamic range, without increasing a device scale and without deterioration in overload noise. SOLUTION: In the case that a transformation coefficient obtained by applying orthogonal transformation to a prediction error signal with a same dynamic range as that of an input signal for each block is quantized and coded, an adaptive quantization device 3 obtains a tentative local decoding signal from a quantized prediction error signal that is obtained by applying inverse orthogonal transformation to a tentative quantization output and compares the tentative local decoding signal with the input signal. In the case that the tentative local decoding signal exceeds a dynamic range, control is conducted by using a conversion signal applying orthogonal transformation to the signal exceeding the dynamic range, and the conversion signal is adaptively quantized and is outputted so that the tentative local decoding signal does not exceed the dynamic range of the input signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a prediction encoding system for digitizing a television signal, quantizing a prediction error signal, compressing and transmitting the data, and an orthogonal transformation system for orthogonally transforming and encoding an image signal. In particular, the present invention relates to a quantization method used when encoding and transmitting a prediction error signal after performing orthogonal transformation and then performing quantization.

[0002]

2. Description of the Related Art Conventionally, a coding system (prediction error orthogonal transform coding system) shown in FIG. 5 is known which combines differential coding (prediction coding) and orthogonal transform coding. As an example of a typical method for orthogonally transforming a prediction error signal, quantizing a transformed signal, and encoding and transmitting the same, for example, TTC standard JT-H261 (published by the Telegraph and Telephone Technical Committee, hereinafter referred to as Reference 1) There is.

Referring to FIG. 5, in this system, a subtractor 4
In step 1, the 8-bit image signal and the 8-bit motion-compensated inter-frame prediction signal are subtracted to obtain a 9-bit difference signal (prediction error signal).
, And performs orthogonal transform by DCT (discrete cosine transform) to obtain 64 types of 12-bit transform coefficients.
Then, the transform coefficient is set to a predetermined quantization characteristic (maximum 1
2-bit dynamic range, range from -2048 to 2
047) is quantized, and the quantized transform coefficients are code-converted by the code converter 44 and transmitted.

[0004] The quantized transform coefficients are also provided to an inverse orthogonal transformer 45, where they are subjected to inverse orthogonal transform and then provided to an adder 46. The adder 46 adds the inverse orthogonal transformer 45 and the prediction signal, and supplies the result to the predictor 47. Then, the predictor 47 generates a motion-compensated inter-frame prediction signal according to the output of the adder 46.

On the other hand, conventionally, an adaptive quantization system for predictive coding is shown in FIG. 6, and this system is described in, for example, JP-A-6-343047.

Referring to FIG. 6, in this system, an 8-bit input signal X is supplied to a subtractor 51 and an adder 52.
The subtracter 51 performs an 8-bit modulo operation, subtracts the prediction signal P from the input signal X to obtain an 8-bit prediction error signal E, and sends the prediction error signal E to the quantizer 53.

[0007] The quantizer 53 has a quantization characteristic of an 8-bit dynamic range, quantizes the prediction error signal E, outputs a temporary quantized output Q, and is one level higher than the temporary quantized output Q. Quantization level Q ₊ and temporary quantization output Q
More One lower quantization level Q _- outputs a.

The subtracter 24 subtracts the provisional quantized output Q and the prediction error signal E to obtain quantization noise qn, and the adder 52 adds the quantization noise qn to the input signal X to add it. A temporary local decoded signal Yt is obtained. Then, the switch 54 determines whether the local decoded signal Yt exceeds or falls below the 8-bit dynamic range, and if it is within the range, outputs the temporary quantized output Q as the quantized output Qs. Further, if it exceeds the range above, the quantization level Q _- outputs as a quantization output Qs, exceeds the range below, and outputs the quantized level Q ₊ as the quantization output Qs. Then, the quantized output Qs is encoded by the code converter 55 and transmitted.

The adder 56 has a quantized output Qs and a prediction signal P
And modulo addition to obtain a local decoded signal Y. Then, the predictor 57 generates the next prediction signal P according to the local decoded signal Y.

On the receiving side, an inverse transform is performed by the code inverse transformer 58 to reproduce a signal having the same quantization level as the quantized output Qs. The quantized signal is modulo-added to the prediction signal by an adder 59 to obtain a locally decoded signal. The predictor 60 generates a prediction signal according to the local decoded signal.

[0011]

By the way, while maintaining the image quality at the same level, the dynamic range of the quantization characteristic is reduced by one.
In order to be able to reduce the coded transmission bit rate by using a quantizer with smaller bits, it is necessary to apply an adaptive quantizer for predictive coding to the quantizer of the above-mentioned prediction error orthogonal transform coding method. In this case, in the prediction error orthogonal transform coding scheme, the transform coding of the prediction error is performed for each 8 × 8 block, and the coding before the quantization is performed.
Later, orthogonal transform / inverse orthogonal transform is performed. Therefore, the temporary local decoded signal of 64 pixels decoded for each block is affected by the quantization noise of 64 transform coefficients for each pixel. For this reason, the quantization error and the pixel signal do not correspond one-to-one, and the determination method in the adaptive quantization method for predictive coding cannot be directly applied to the determination method for prediction error orthogonal transform coding.

When a pixel signal exceeds a dynamic range in one block, it is assumed that a simple brute force method is used to determine which transform coefficient exceeds the dynamic range due to the influence of quantization noise. , 6
It is necessary to find the effect of quantization noise on a combination of four types of transform coefficients. In the worst case, the amount of calculation for brute force is enormous, and it is not practical to implement a decision circuit to perform in real time. There is a problem.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a quantization characteristic capable of encoding an image having an image quality equivalent to that of the conventional quantization without deterioration of overload noise with a quantization characteristic having a smaller dynamic range by one bit than the conventional one without increasing the apparatus scale. It is to provide a generalization method.

Another object of the present invention is to provide a quantization system capable of performing transmission with a smaller number of quantization levels (number of allocated bits), that is, a smaller number of coded transmission bits, as compared with the related art. .

[0015]

When the adaptive quantizer for predictive coding is directly applied to the predictive error orthogonal transform coding method, a transform coefficient which greatly affects a pixel signal exceeding a dynamic range is obtained. The adaptive quantization is determined by a method of checking whether the provisional quantized output level of the transform coefficient can be reduced by one so as to prevent the provisional local decoded signal from exceeding the dynamic range.

The transform coefficient that greatly affects the dynamic range over of the decoded signal can be estimated from the amplitude value of the transform coefficient of the out-of-range signal and the magnitude of the quantization noise of the quantized output.

Here, as the out-of-range signal, a pixel point within the dynamic range is 0, and a pixel exceeding the upper limit is defined as a positive amplitude value of the exceeded portion as an amplitude value of the pixel point, , The signal value of one block is obtained by using the negative amplitude value of the excess portion as the amplitude value of the pixel point.

Next, the out-of-range signal is subjected to inverse orthogonal transform to determine the amplitude value of the converted signal. Assuming that a transform coefficient having a large amplitude is greatly affected by quantization noise, the quantization step is reduced by one or By increasing the value, it is determined whether the temporary local decoded signal does not exceed the dynamic range. If the dynamic range is exceeded, the determination is repeated again. By this procedure, by forcibly changing the quantization level of the transform coefficient having a large influence, the quantization noise serves to correct the pixel signal having an overrange in the opposite direction. The signal does not exceed the dynamic range, and by this determination method, it can be quickly determined not to exceed the dynamic range.

According to the configuration of the present invention, the orthogonal transform signal of the prediction error signal can be quantized and coded and transmitted with the quantization characteristic of the dynamic range smaller by one bit than the conventional one, and the reproduced image is equivalent to the conventional one without overload. Can be decoded.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings.

Referring to FIG. 1, an 8-bit digital input signal X (−128 to 127) is supplied to an 8-bit dynamic range subtracter 1, and the subtractor 1 outputs a prediction signal P from a predictor 7. And an input signal X are subtracted by a modulo operation to obtain an 8-bit prediction error signal E, which is supplied to the orthogonal transformer 2.

In the orthogonal transformer 2, “8 samples” × “8
The prediction error signal E is orthogonally transformed for each block using 64 pixels (Eij = E11 to E88) of the "line" as one block, and 64 types of transform coefficients (Et11 to Et88) are output as one block. The 64 types of transform coefficients have a dynamic range of 11 bits and are supplied to the adaptive quantizer 3.

The adaptive quantizer 3 includes a quantizer 10 and a selector 1
1, inverse orthogonal transformer 12, adder 13, range determiner 1
5, an orthogonal transformer 16, a control determiner 17, and an output switch (SW) 18. The quantizer 10 has an 11-bit dynamic range, quantizes each transform coefficient for each block with quantization characteristics corresponding to each transform coefficient, and supplies each temporary quantized output Q to the selector 11. Further, the quantizer 10 is one above quantization output Qij of the temporary (Q ⁺⁾ and one lower ^- supplying level quantized output even together and output to the selector 11 of the (Q).

The selector 11 is controlled determiner provisional in response to the control signal from 17 Qij a quantized output, Q ^+, and Q ^- provisionally for each block to select one of the quantization output Qt And supplies it to the inverse orthogonal transformer 12 and the output SW 18.

The inverse orthogonal transformer 12 has an inverse transform characteristic of the orthogonal transformer 2, and performs an inverse orthogonal transform of the temporary quantized output Qt for each block to thereby temporarily quantize a prediction error signal having a dynamic range of 8 bits. Eq is reproduced and supplied to the adder 13.

The adder 13 adds the prediction signal P supplied from the predictor 7 and the provisional quantized prediction error signal Eq by an 8-bit modulo operation to determine a provisional 8-bit local decoded signal Y. It is supplied to the range determiner 15.

The range determiner 15 first compares the input signal X with the temporary local decoded signal Y to determine whether the local decoded signal Y exceeds the dynamic range. For comparison, a noise signal is obtained by subtracting the temporary local decoded signal Y from the input signal X by a 9-bit subtractor (not shown). When the noise signal N is larger than half of the dynamic range, the provisional local decoded signal Y overflows or underflows the 8-bit dynamic range due to the influence of quantization noise, and a correct local decoded signal is not obtained. And a value exceeding the range boundary of each pixel as an out-of-range signal is obtained for each block. The out-of-range signal has a positive value in the case of an overflow and a negative value in the case of an underflow. If the pixel is determined to be within the range, the out-of-range signal has a value of 0.
The thus-obtained out-of-range signal composed of 64 pixels is supplied to the orthogonal transformer 16 and the control determiner 17.

The orthogonal transformer 16 has the same conversion characteristics as the orthogonal transformer 2, and performs orthogonal transform on the out-of-range signal for each block to obtain a transform coefficient and supplies it to the control decision unit 17. First, the control determiner 17 performs control so that the temporary quantized output Q output from the quantizer 10 is selected. If the out-of-range signals supplied to the control determiner 17 are all 0 in the first comparison determination, the control determiner 17
Sends a control signal to the output SW 18 so that the provisional quantized output Q output from the selector 11 is output as the finally selected quantized output.

If the out-of-range signal is not 0, the quantization output Q ⁺ or Q ^- is selected as the level of the provisional quantization output corresponding to the largest conversion coefficient of the out-of-range signal. A control signal is sent to the device 11. As a result,
The above-described comparison determination control process is repeated for the newly output temporary quantized output.

After the second time, this comparison, determination and control process is repeated until the range signal becomes 0 for all blocks.

As a method of controlling the quantized output based on the transform coefficient, a method of performing the quantization adaptive control using only the transform coefficient of the DC component will be described.

The quantization noise N (f (x, y)) obtained by inversely transforming the quantization noise Nt of the transform coefficient is given by the following equation (1), and the transform coefficient quantization noise Nt (F (u, v)) ) Is multiplied by a factor of 1/8. As a result, when the overflow occurs, the quantization level of the DC component of the transform coefficient is decreased by one, and when the underflow occurs, the quantization level is increased by one.

When the magnitude of the out-of-range signal is small, a complicated judgment can be made only by correcting the DC component, and correction can be easily made so as not to exceed the range. When the amplitude is large, a factor of 1/8 is applied, so that a change of one level does not make the correction effective, and correction of two or more levels may be required. However, the DC quantization noise becomes more noticeable as block distortion,
When the amplitude is large, it is necessary to control the selection of the quantization level of the temporary quantization output including the transform coefficient of the AC component.

The output SW 18 outputs the quantized output supplied from the selector 11 and supplies it to the code converter 4 and the inverse orthogonal transformer 5. The code converter 4 codes-converts each level of the quantized output into a transmission code for each block, multiplexes it with other information, and sends it to the transmission path. The inverse orthogonal transformer 5 has the inverse transform characteristic of the orthogonal transformer 2, inversely transforms the quantized output, obtains a quantized prediction error signal Eq, and supplies it to the adder 6.

The adder 6 has an 8-bit dynamic range, modulo-adds the quantized prediction error signal Eq and the prediction conversion coefficient P to obtain an 8-bit local decoded signal Y, and supplies it to the predictor 7. I do. The predictor 7 calculates and outputs the next predicted signal P from the local decoded signal Y for each block according to the prediction characteristics, and supplies the same to the subtractor 1, the adder 6, and the adder 13.

The predictor 7 has a function of performing motion compensation prediction, finds an optimal motion vector for the next input block for each block by a matching method, and outputs a motion-corrected prediction signal P. The motion vector is encoded together with the quantized output signal and sent to the receiving side.

Next, the characteristics of the orthogonal transformers 1 and 16 and the inverse orthogonal transformers 5 and 12 will be described.

In the DCT transform of 8 rows and 8 columns in the orthogonal transformer, a two-dimensional discrete cosine transform that can be separated into an 8 × 8 one-dimensional transform is performed. The signal of one block of 8 rows and 8 columns is represented by f
(x, y) (corresponding to X11 to X88) and the conversion output coefficient of 8 rows and 8 columns as F (u, v) (corresponding to Xt11 to Xt88), the conversion output F (u, v) becomes TTC The one given by Equation 1 shown in Recommendation H.261 is used.

[0039]

(Equation 1) Note that for the block to be converted, x = 0 corresponds to the left end of the block, and y = 0 corresponds to the upper end of the block.

On the other hand, the inverse transform characteristic of the inverse orthogonal transformer is as shown in Equation 2.

[0041]

(Equation 2) When this conversion is performed, the conversion coefficient F becomes a signal whose dynamic range is widened by 3 bits. That is, when the signal f has 8 bits, the conversion coefficient F has a dynamic range of 11 bits.

In the conventional difference coding, the difference signal E obtained by subtracting the 8-bit prediction signal P from the 8-bit input signal X is 9 bits. Is a 12-bit dynamic range. In other words, in the conventional example, the difference conversion coefficient is 12
It becomes a dynamic range of bits. To quantize this transform coefficient, a quantization characteristic requires a dynamic range of 12 bits.

On the other hand, in the present invention, the 8-bit input signal X
, The 8-bit prediction signal P is subtracted by a bit modulo operation to obtain an 8-bit difference signal (prediction error signal). The transform coefficient obtained by orthogonally transforming the 8-bit difference signal is 11
It is sufficient that the quantizer 10 quantizes the signal of the conversion coefficient of the 11-bit dynamic range, and the dynamic range of the quantization characteristic is a half (1 bit less) 11-bit range as compared with the conventional one. That would be fine.

Next, the operation of the range determiner will be described.

In the predictive coding, if the input signal is X, the local decoded signal is Y, and the quantization noise is N, Y = X × N
There is a relationship. In the present invention, the prediction error signal is obtained by a modulo arithmetic unit having the same dynamic range as the input signal (n = 8 bits), and the local decoded signal is also obtained by modulo addition of the same dynamic range. For this reason, the locally decoded signal is represented by Ym = MOD (X + N, n (value modulo n bits), from -2 ** (n-1) to 2 ** (n-
1) The value is in the range of -1.

Therefore, for example, X is the upper limit value of the range T =
When the quantization noise N is larger than XT-X when the value is close to 2 ** (n-1) -1, the local decoded signal Y = X + N overflows and exceeds the dynamic range, and the actual local decoded signal Is Ym = MOD (X + N, n).
= X + N-2 ** n, and the lower limit B = -2 ** (n-
This is a negative value close to 1).

Similarly, when an underflow occurs due to the quantization noise N, a locally decoded signal that overflows due to the influence of the quantization noise on an input signal near the lower limit similarly takes a value near the upper limit.

Whether the overflow or underflow has occurred is determined by the noise N obtained by subtracting the local decoded signal Ym from the input signal X.
Can be determined from The quantization noise of the quantizer is usually 2 **
Since it is set smaller than half of (n-1), overflow occurs when X-Ym> 2 ** (n-1). Underflow occurs when X-Ym <-2 ** (n-1). The correct local decoded signal is obtained within the range.

A simple determination is made by subtracting the (n + 1) -bit subtraction value (X
-Ym) is determined by looking at the upper two bits. If the upper two bits are "01", an overflow occurs, if "10", an underflow occurs, and if "11" or "00", a correct local decoded signal is obtained.

As described above, the range determiner 15 compares and determines the range of the signal, and obtains and outputs the out-of-range signal.
With respect to the local decoded signal exceeding the dynamic range due to the influence of quantization noise, a transform coefficient affecting the signal exceeding the range is obtained. Then, in order to prevent the local decoded signal from exceeding the dynamic range due to the influence of the quantization noise of the quantization output of the transform coefficient, the selected quantization level is replaced by one level above or one level below the selected level. Is selected and output.

The out-of-range signal indicating the out-of-range signal is obtained for each block, with the signal value not exceeding the range being 0 and the signal exceeding the range being the exceeded value. The out-of-range signal has a positive value in the case of overflow and a negative value in the case of underflow.

Next, an example of the operation of the control determiner will be described.

In the case where the local decoded signal of one block exceeds dynamic, it is determined which transform coefficient has a large influence of quantization noise. An out-of-range signal indicating a part of the signal exceeding the dynamic range is orthogonally transformed for each block to obtain a transform coefficient. Among the obtained conversion coefficients, the one having the largest amplitude is obtained.

In the case of overflow, when the transform coefficient having the largest amplitude is a positive value, control is performed so that the next lower quantization level is selected and output instead of the temporary quantized output corresponding to the transform coefficient. When the value is a negative value, control is performed so as to select the next higher quantization level instead of the provisional quantization output corresponding to the transform coefficient.

In the case of underflow, when the transform coefficient having the largest amplitude is a positive value, control is performed so as to select and output the next higher quantization level instead of the provisional quantized output corresponding to the transform coefficient. When the value is a negative value, control is performed so as to select the next lower quantization level instead of the provisional quantization output corresponding to the transform coefficient.

Another example of the operation of the control determiner will be described.

If the dynamic range of the local decoded signal of one block exceeds the dynamic range, it is determined which transform coefficient has a large influence of quantization noise. An out-of-range signal indicating a part of the signal exceeding the dynamic range is orthogonally transformed for each block to obtain a transform coefficient. Among the obtained conversion coefficients, the one having the largest amplitude is obtained.

If an overflow has occurred, a positive conversion coefficient having the largest amplitude is searched for. Control is performed so as to select and output the next lower quantization level instead of the provisional quantization output corresponding to the transform coefficient.

The reason why the conversion coefficient is limited to the positive conversion coefficient is that, when the conversion coefficient is a negative value, the control is performed so as to select the next higher level instead of the temporary quantization output corresponding to the conversion coefficient. Coefficient quantization noise will have a positive value added, and for components close to DC, quantization noise will be added in a more positive direction for pixels, possibly preventing overflow. This is because they can no longer do it.

When an underflow has occurred, a conversion coefficient having the largest amplitude among positive conversion coefficients is searched for. Control is performed such that the next higher quantization level is selected and output instead of the temporary quantization output corresponding to the transform coefficient. The reason why the conversion coefficient is limited to the positive conversion coefficient is that when the conversion coefficient is a negative value, if the control is performed so that the next lower level is selected instead of the temporary quantization output corresponding to the conversion coefficient, the quantization of the conversion coefficient becomes Negative values will be added to the quantization noise, and for components close to DC, quantization noise will be added more negatively to the pixel, and in some cases, underflow can be prevented. It is because it disappears.

With reference to FIG. 2, a second example according to the present invention will be described.

In the example shown in FIG. 1, a temporary local decoded signal is obtained by the adaptive quantizer 3. In this example, the quantized prediction error signal Eq is subtracted from the prediction error signal to perform quantization. The noise N is obtained (in the illustrated example, the same components as those in the example illustrated in FIG. 1 are denoted by the same reference numerals).

The subtracter 21 subtracts the quantized prediction error signal from the prediction error signal to obtain quantization noise N, and supplies it to the range determiner 22. In the range determiner 22, when the input signal X and the quantization noise N are added by an 8-bit adder,
A temporary local decoded signal Y is obtained.

The comparison of whether or not the dynamic range has been exceeded is performed by a 9-bit adder using the input signal X and the quantization noise N.
And looking at the upper 2 bits of the 9-bit output,
It can be determined whether or not the temporary local decoded signal exceeds the dynamic range. That is, the input signal X (−128 to
127), when the quantization noise N is added to exceed the dynamic range of 8 bits, an overflow or an underflow occurs, and the upper 2 bits are “01”.
In the case of (128 or more), overflow has occurred,
In the case of “10” (−129 or less), it means that an underflow has occurred.

Referring to FIG. 3, another specific example of the adaptive quantizer will be described.

In the illustrated example, the same components as those of the adaptive quantizer shown in FIG. 1 are denoted by the same reference numerals.
In the illustrated example, the decision controller 32 performs control in consideration of the magnitude of the quantization noise Nt added to the transform coefficient by the quantizer 10 in addition to the transform coefficient of the out-of-range signal.

The subtractor 31 subtracts the quantized transform coefficient from the transform coefficient supplied from the orthogonal transformer 2 to obtain quantization noise, and supplies it to the decision controller 32. Judgment controller 32
Is the transform coefficient of the out-of-range signal and the quantization noise N of the transform coefficient.
Based on the information of t, selection control of the quantization step is performed. For 64 types of transform coefficients, it is determined that the quantization noise due to the coefficient whose product value is large has a large effect on the overflow or underflow of the temporary local decoded signal, and the provisional quantum of the corresponding transform coefficient is determined. Control to select the next higher or lower quantization level instead of the quantization output.

If the corresponding quantization noise is written and stored in a ROM (read-only memory) in advance, a subtractor is unnecessary only with the ROM, and if shared with the ROM of the quantizer, H
/ W configuration is simplified.

Referring to FIG. 4, another specific example of the adaptive quantizer will be described.

Although the illustrated example is substantially the same as the configuration shown in FIG. 3, in this example, in the subtractor 31, the temporary coefficient output from the selector 11 is output from the transform coefficient supplied from the orthogonal transformer 2. The quantization noise is obtained by subtracting the quantization output. When the alternative provisional quantization level is selected by the selector 11 by the control signal, the provisional quantization output changes. According to the configuration of the illustrated example, the selected provisional quantization output is obtained, and accurate transform coefficient quantization noise is obtained.

The quantization characteristics may be prepared separately for each transform coefficient, but the control becomes complicated. For example, there is a method of dividing the quantization characteristic for DC component for intra-frame encoding and the other transform coefficients. Then, the same quantization characteristic is applied in a predetermined range (macroblock).

The following method is used as another method. The high-frequency component, in which the component of the transform coefficient is small, is visually coarsened and quantized to make it inconspicuous, and the gain is increased by a bit shift or the like in advance according to the transform coefficient, and then quantized with the same quantization characteristic. . Thereby, the transform coefficient of the high frequency component is roughly quantized equivalently.

In the case of performing each quantization level variable length coding (entropy coding), if a variable length code corresponding to the frequency of occurrence of each level is used, efficient coding can be performed even if the number of levels increases. Although possible, the limitation on the number of quantization levels is necessary for efficient coding when a large amount of information is generated in a scene change and the amount of generated information is too large.

[0074]

As described above, according to the present invention,
Since the transform coefficient of the orthogonal transform can be quantized using half the quantization characteristic in which the dynamic range of the quantizer is one bit smaller than the conventional one, the number of bits for coding the quantized output can be reduced, and the efficiency can be reduced. There is an effect that efficient encoded transmission can be performed.

Further, according to the present invention, when orthogonally transforming the prediction difference signal of the predictive coding method to quantize the transform coefficient and transmit the coded signal, the dynamic range of the quantization characteristic can be improved with a small number of bits, and the transient response characteristic can be improved. Can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram for explaining a first example of an adaptive quantization scheme for orthogonal transform encoding according to the present invention.

FIG. 2 is a block diagram for explaining a second example of the adaptive quantization scheme for orthogonal transform coding according to the present invention.

FIG. 3 is a block diagram showing another example of the adaptive quantizer shown in FIG. 1;

FIG. 4 is a block diagram showing still another example of the adaptive quantizer shown in FIG. 1;

FIG. 5 is a block diagram for explaining conventional inter-frame prediction orthogonal transform coding.

FIG. 6 is a block diagram for explaining DPCM (prediction coding) using conventional adaptive quantization.

[Explanation of symbols]

 1,21 subtractor 2 orthogonal transformer 3 adaptive quantizer 4 code transformer 7 predictor 10 quantizer 11 selector 12,5 inverse orthogonal transformer 13,6 adder 15,22 range determiner 16 orthogonal transformer 17, 32 control decision unit 18 output SW

Claims (7)

[Claims] 1. When quantizing and encoding a transform coefficient obtained by orthogonally transforming a prediction error signal having the same dynamic range as an input signal for each block, the transform coefficient is quantized for each block and a temporary Obtaining a quantized output, obtaining a temporary local decoded signal in accordance with a quantization prediction error signal obtained by performing an inverse orthogonal transform on the temporary quantized output, comparing the temporary local decoded signal with the input signal, Output means for adaptively quantizing the transform coefficient according to a transform signal obtained by orthogonally transforming the signal exceeding the dynamic range when the temporary local decoded signal exceeds the dynamic range, and outputting the quantized output as a quantized output; An adaptive quantization method for orthogonal transform coding, characterized in that: 2. The adaptive quantization scheme for orthogonal transform coding according to claim 1, wherein said output means includes a prediction signal obtained based on said quantization prediction error signal and said provisional quantization output, and An adaptive quantization method for orthogonal transform coding, wherein the provisional local decoded signal is obtained by modulo addition in a dynamic range. 3. The adaptive quantization method for orthogonal transform coding according to claim 1, wherein said output means obtains quantization noise by subtracting said quantization prediction error signal from said prediction error signal. An adaptive quantization scheme for orthogonal transform coding, wherein the temporary local decoded signal is obtained by adding the input signal to noise. 4. The adaptive quantization scheme for orthogonal transform coding according to claim 1, wherein said output means obtains said temporary local decoded signal in accordance with said quantized prediction error signal, and outputs said temporary local decoded signal and said input signal. Comparing with the temporary local decoded signal, if the temporary local decoded signal exceeds the dynamic range, the quantization level next to the temporary quantized output is changed instead of the temporary quantized output. First means for outputting as an alternative temporary quantized output, and second means for repeatedly executing the first means to obtain a temporary quantized output in which the temporary local decoded signal does not exceed the dynamic range. An adaptive quantization scheme for orthogonal transform coding, characterized by having: 5. The adaptive quantization method for orthogonal transform encoding according to claim 1, wherein the temporary local decoded signal is obtained from a signal obtained by performing an inverse orthogonal transform on the temporary quantized output, and the input signal and the temporary signal are obtained. If the provisional local decoded signal exceeds the dynamic range as compared with the provisional local decoded signal, a transforming signal obtained by orthogonally transforming the signal exceeding the dynamic range and the provisional quantized output are generated. A first means for outputting a quantization level adjacent to the provisional quantization output as a replacement provisional quantization output in place of the provisional quantization output in accordance with quantization noise; and A second means for repeatedly obtaining a temporary quantized output in which the temporary local decoded signal does not exceed the dynamic range, wherein an adaptive quantization scheme for orthogonal transform coding is provided. 6. When quantizing and encoding a transform coefficient obtained by orthogonally transforming a prediction error signal having the same dynamic range as an input signal for each block, the prediction signal is modulo-modulated from the input signal in the dynamic range. First means for subtracting to obtain the prediction error signal, second means for orthogonally transforming the prediction error signal for each block and outputting the transform coefficient, and The transform coefficients are quantized for each block to obtain a temporary quantized output, and a temporary local decoded signal is obtained by using a transform signal obtained by performing an inverse orthogonal transform on the temporary quantized output to obtain the input signal and the temporary signal. If the provisional local decoded signal exceeds the dynamic range as compared with the local decoded signal of the above, according to the converted signal obtained by orthogonally transforming the signal of the portion exceeding the dynamic range. As a result of performing the selection control, instead of the provisional quantization output corresponding to the selected transform coefficient, the next higher or lower quantization level is output as a replacement provisional quantization output, and the provisional local decoded signal is output. Third means for performing the selection control to obtain a quantized output until it falls within the dynamic range, code converting means for converting the quantized output into a transmission line code, and performing inverse orthogonal transform on the quantized output. A fourth means for obtaining a quantized prediction error signal; a fifth means for adding the prediction signal and the quantized prediction error signal by a modulo operation to obtain a local decoded signal; 5. An adaptive quantization scheme for orthogonal transform coding, comprising: a fifth means for obtaining a prediction signal. 7. The adaptive quantization method for orthogonal transform coding according to claim 6, wherein said third means performs the selection control and further includes a magnitude of quantization noise of said transform signal in addition to said transform signal. An adaptive quantization scheme for orthogonal transform coding, characterized by using information representing