JPH02312383A - Encoding and decoding device - Google Patents

Abstract

PURPOSE:To prevent a picture quality from deteriorating even if a quantization characteristic is set to be non-linear by selectively using a first quantizer which quantizes a differential value between an input sampled value and the prediction value and a second quantizer which quantizes the input sampled value in accordance with the change of the input sampled value and executing encoding. CONSTITUTION:The first quantizer 36 which quantizes the differential value between the input sampled value Xi and the prediction value, and the second quantizer 36 which quantizes the input sampled value Xi are provided. An encoding code is obtained by selectively using the first and the second quantizers 36 and 34 in accordance with the change of the input sampled value Xi. Thus, an encoding device which can reduce the deterioration of the edge part and the flat part of the sampled value even if the quantization characteristic is set to be non-linear can be obtained.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention provides an encoding device for encoding a difference value between a sample value and its predicted value, and for decoding an encoded code obtained by encoding the difference value. This invention relates to a decoding device.

[Conventional technology]

When digitizing and transmitting image information, audio information, etc., encoding is performed to reduce the amount of information to be transmitted in order to save on transmission costs. Furthermore, decoding is performed to restore the encoded and transmitted information to its original state.

One of the encoding methods is differential encoding (Differential encoding).
alPulse Code Modulation
(hereinafter referred to as DPCM), which compresses the amount of information by utilizing the correlation between adjacent sample values. Specifically, the coded sample value is decoded once, the decoded value (locally decoded value) is used to find the predicted value for the next sample value, and the difference between this predicted value and the input sample value is quantized. A method of encoding is used.

FIG. 11 is a block diagram showing the configuration of a conventional encoding device that performs the simplest prior value predictive encoding.

In FIG. 11, lO is the sample value X to be encoded
Input terminal where I is input, 12 is sample value XI and predicted value P
14 is a quantizer that quantizes the difference value from the subtracter 12, 16 is an output terminal that outputs the DPCM code Y1 from the quantizer 14, and 18 is the quantizer 14 20 is a D-type flip-flop 2;
22 is a limiter that limits the amplitude of the locally decoded value X1 output from the adder 20 within the dynamic range of the input sample value; 24 is the limiter 22; The subtracter 1 uses the local decoded value X+ outputted from the subtracter 1 as the predicted value P1+1 of the next sample value X +++.
This is a D-type flip-flop that applies voltage to 2.

Next, the operation of the configuration shown in FIG. 11 will be explained.

The sample value X inputted to the input terminal 10 is applied to a subtracter 12, and the predicted value P+ is subtracted by the subtracter 12. The difference value output from the subtracter 12 is quantized by the quantizer 14, and is sent to the output terminal 1 as a DPCM code Yl.
6 is output. D P CM output from the quantizer 14
Code Y! is also input to the inverse quantizer 18, where DP
After the CM code Y1 is decoded into a quantized representative value of the difference value, it is output to the adder 20.

The adder 20 adds D to the quantized representative value from the inverse quantizer 18.
Previous value predicted value P1 output from type flip-flop 24
are added, thereby restoring the difference value (quantized representative value) to the sample value (locally decoded value). The restored sample value (locally decoded value) is applied to a D-type flip-flop 24 after its amplitude is limited to a predetermined range by a limiter 22 . The D-type flip-flop 24 outputs the locally decoded value output from the limiter 22 in synchronization with the clock, and applies it to the subtracter 12 and the adder 20 as a predicted value P +++ for the next sample value.

Generally, the distribution of difference values of predicted values is biased toward small values, and by encoding the difference values using nonlinear quantization and transmitting them, it is possible to reduce the amount of information.

[Problem that the invention seeks to solve]

In the conventional encoding and decoding devices as described above, the range of values that the input sample value can take, that is, the dynamic range (hereinafter referred to as D range) is O to (n-1), and the width is n levels. , the D range of the difference value is (-n+1) ~
(n-1), and its width is (2n-1) levels, which is approximately twice the width of the D range of the input sample value. In this specification, for convenience of explanation, the range of possible values is expressed as D range,
The range of possible values is expressed by dividing it into D Reno 2 width. As a result, when the quantization characteristic is a nonlinear characteristic, the maximum value of the quantization error (difference between the difference value before quantization and the difference value after quantization) far from the predicted value becomes a very large value. This was a major cause of image quality deterioration (edge business) at edge portions with large difference values.

On the other hand, if the quantization characteristics are brought closer to linear characteristics in order to reduce the above-mentioned deterioration, the quantization error near the predicted value will increase, causing roughness in the reproduced image (increase in granular noise) in flat areas, and In places where the level changes gradually, such as a face, contours that look like contour lines on a map (false contours) occur, which degrades the image quality.

Furthermore, in the conventional example, when an error occurs in the transmission path, the decoded value obtained by the decoding device becomes a different value from the local decoded value obtained by the encoder, and this erroneous decoded value becomes the predicted value of the next decoded value. As a result, errors propagate one after another, making it impossible to restore the original image.

This invention was made to solve such problems, and even when the quantization characteristics are made non-linear, it reduces the deterioration of the edge portions and flat portions of sample values, and also reduces error propagation of decoded values. The purpose is to provide an encoding device and a decoding device that can be realized.

[Means for solving problems]

In order to achieve the above object, the encoding device of the present invention includes:
a first encoding means for quantizing a difference value between an input sample value and its predicted value; and a second encoding means for quantizing the input sample value; The encoded code is obtained by selectively using the eleventh second encoding means.

Further, the decoding device of the present invention includes a first encoding means for quantizing a difference value between an input sample value and its predicted value, and a second encoding means for quantizing the input sample value. A decoding device that decodes an encoded code obtained by selectively using and encoding a value according to a change in value, the first. The first . comprising second decoding means;
Part 1. The second decoding means is selectively used to obtain a decoded value.

[Effect]

By configuring as described above, it is possible to use the first encoding means having a sufficiently fine quantization step in a range where the absolute value of the difference value is small in a flat part or a part where the value changes slowly. Since quantization can be performed in fine quantization steps, it is possible to prevent the occurrence of granular noise, etc. In addition, the edge portion can use a second encoding means to quantize the sample value, and the influence of gradient overload can also be reduced. .Moreover, the second
Error propagation is cleared every time the encoding means is used, and the adverse effects of error propagation can also be reduced.

Furthermore, in the decoding device of the present invention, the first decoding means can perform decoding in sufficiently fine steps in a flat area, and granular noise, false contours, etc. do not occur (also,
The influence of gradient overload on edge portions and image quality deterioration due to error propagation can also be reduced by using the second decoding means.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an encoding device according to an embodiment of the present invention. In the figure, 30 is an input terminal for the sample value X+, 32 is a subtracter, 34.36 is a quantizer, 38.50 is a switch, 4
0 is the output terminal of the encoded code Y1, 42° 44 is the inverse quantizer, 46 is the adder, 48 is the limiter, 52.56 is D
The type flip-flop 54 is a range end detection circuit which will be described later.

A subtracter 32 calculates a difference between the sample value XI input from the input terminal 30 and the predicted value XI, which is applied to a quantizer B36. The quantizer B36 quantizes the applied difference value and applies the DPCM code Ys+ to the terminal of the switch 38 and the inverse quantizer 44. The inverse quantizer 44 is a DPCM
The code is converted into a quantized representative value (that is, a difference value after quantization) and applied to the adder 46.

The adder 46 adds the quantized difference value to the predicted value X1 applied from the D-type flip-flop 52 to restore the original input sample value. The restored input sample area is the limiter 48
The local decoded value X a+ is limited to a value within the range of the input sample value and is applied to the terminal of the switch 50 .

On the other hand, the sample value x1 input from the input terminal is sent to the quantizer A.
34 and is quantized into a PCM code YAI, which is applied to terminal a of the switch 38 and the inverse quantizer 42. The inverse quantizer 42 restores the quantized representative value, that is, the original input sample value, from the PCM code YAI, and applies it to the terminal a of the switch 50 as the local restored value XhI. In the switch 50, according to the range end detection signal L+-+ of the previous pixel applied from the D-type flip-flop 56, Li-1
is at a high level (H'), the local decoded value X of terminal a
At is selected, and when Ll-+ is at low level (“L”), the locally decoded value XBI of the terminal S is selected and applied to the D-type flip-flop 52. The D-type flip-flop 52 applies the locally decoded value X+ applied from the switch 50 to the subtracter 32 as a predicted value XI+1 in the next clock cycle.

In the switch 38, like the switch 50 described above, Ll-
When 1 is “H”, the PCM code YAI of terminal a is
When +-+ is L”, the DPCM code of the terminal is Y a
t is selected and applied to the output terminal 40 and the range end detection circuit 54 as the encoded code Yl.

The range end detection circuit 54 detects whether the DPCM code YBI is a code indicating the range end of the quantized difference value, that is, the maximum value or minimum value of the quantized representative value, and if it is a code indicating the range end, " A detection signal L+, which is H" and otherwise L, is applied to the D-type flip-flop 56. In an actual circuit, in order to facilitate this range end detection,
If the DPCM code is 4 bits, the code indicating the maximum value of the quantization representative value is 111 bits, and the code indicating the minimum value is o.
Assign a specific code such as "ooo", and use an AND circuit and N
Detection is performed using an OR circuit or the like.

The D-type flip-flop 56 transfers the range end detection signal to the switch 38 . 50 and the range end detection circuit 54. The reason why this range end detection circuit signal is also applied to the range end detection circuit 54 is to prevent range end detection from being performed when the encoded code is the PCM code YAM.

Next, the operation of the encoding device of the embodiment will be explained.

As described above, the probability of appearance of the difference value between the predicted value X and the input sample value XI is concentrated near 0, and therefore the switch 38.50 normally selects the DPCM code Y 13i of the terminal. That is, DPCM encoding is performed in which the D range of the quantization representative value is narrowly limited.

By the way, if the number of quantization levels is constant, the narrower the D range of the quantization representative value, the finer the quantization step (the finer the quantization step), the more problems such as granular noise, false contours, edge business, etc. will be improved. On the other hand, the quantization representative value D
As the range becomes smaller, the followability of decoded edges deteriorates, resulting in gradient overload that blunts the edges. For example, if the input sample value range is O to 255 and the quantization representative range is -63 to 63, and the input value changes stepwise from O to 255, the decoded value changes from O to 255. It takes more than 4 pixels.

In this embodiment, in pixels at the edge where gradient overload occurs, DPCM'[ is not processed by the child converter B36 but by the
By using the M quantizer A34, the deterioration caused by the above gradient overload is suppressed.

Furthermore, the quantization characteristic of the PCM quantizer A34 may be approximately linear because the correlation with the previous pixel is low in the edge portion and the difference sensitivity is low in human visual characteristics. Therefore, quantizer A34 and inverse quantizer A
42 can be constructed by bit shifting, which greatly simplifies the hardware. For example, input sample value 8 bits,
When the encoding code is 4 bits, the quantizer A34 only needs to output the upper 4 bits, and the inverse quantizer A42
Then, the upper 4 bits may output the encoded code, and the lower 4 bits may output 0111'', which is the median value of the quantization steps.

Further, switching of the DPCMPCM quantizer A34M quantizer A34 may be performed by transmitting a signal for switching, but in this embodiment, in order to save the amount of information to be transmitted, the determination is made using an encoded code. , switch the quantizer. That is, D
If the encoded code quantized by the PCM quantizer 836 indicates a value at the end of the range of the D range limited quantization representative value, it is determined that it is an edge, and the next pixel is PCM quantized. be.

Therefore, when the step waveform that rapidly changes from 0 to 255 is input, the decoded value changes from 0 to 63 to 248, and is followed by two pixels.

Next, a decoding device according to an embodiment of the present invention will be explained.

FIG. 2 is a block diagram showing a decoding device according to an embodiment of the present invention. In the figure, 60 is an input terminal for the encoded code Y;
62, 64 are inverse quantizers A and B, 66 is an adder, 68 is a limiter, 70. 72 is a switch, 74 is a decoded value output terminal, 76.80.84 is a D-type flip-flop, 7
8 is a range end detection circuit, 82 is a comparator, and 86 is a decoding position replacement determination circuit.

The encoded code Y1 input from the input terminal 60 is applied to the inverse quantizers 62 and 64 and the range end detection circuit 7. Inverse quantizer A62. B64 are inverse quantizers A42 .B64 of FIG. 1, respectively. This circuit has the same function as B44, and by applying the encoding code Y+, the dequantizer A62 sends the quantized input sample value to the terminal a of the switch 70, and the dequantizer B64 sends the quantized input sample value to the terminal a of the switch 70. The difference values are applied to adders 66, respectively. The adder 66 adds the predicted value X1 applied from the D-type flip-flop 76 to the difference value. The output limiter 48 of the adder 66 limits the input sample value to a value within the D range and applies it to the terminal of the switch 70.

The switch 70 switches Li according to the range end detection signal L+-+ of the previous pixel applied from the D-type flip-flop
When −+ is “L”, outputs the decoded value XAI of terminal a,
When L+-+ is "L", the decoded value 5'Bi of the terminal S is outputted and applied to the terminal C of the switch 72, the D-type flip-flop 76, and the comparator 82.

The D-type flip-flop 76 converts the decoded value applied from the switch 70 into the predicted value XI+1 in the next clock cycle.
It is applied to the adder 66, the comparator 82, and the terminal d of the switch 72 as a signal. The switch 72 selects h in accordance with a replacement determination signal 11 applied from a decoding position replacement determination circuit 86 to be described later.
When J1="H", the signal at terminal C is output, and when J1="L", the signal at terminal d is selected.The output of switch 72 is output from output terminal 74 as a decoded value.

The range end detection circuit 78 has the same function as the range end detection circuit 54 in the encoding device shown in FIG. The detection signal L1 is set to "H", and otherwise the detection signal L1 is set to "L".

The D-type flip-flop 80 applies this range end detection signal LI to the switch 70, the range end detection circuit 78, and the decoding position replacement determination circuit 86 in the next clock cycle.

The comparator 82 compares the predicted value X1 and the decoded value X1, and applies only the sign M1 indicating the sign of the difference value to the D-type flip-flop 84 and the replacement determination circuit 86. The D-type flip-flop 84 applies the difference value sign M1 to the decoding position replacement determination circuit 86 in the next clock cycle.

The decoding position replacement determination circuit 86 receives the range end determination signal L+-+
is "H" and the difference value sign M+ matches the difference value sign M+-+ of the previous pixel, it is determined that gradient overload has occurred, and the decoded value replacement judgment signal! + is set to H", and otherwise the replacement determination signal It is set to "L".

Therefore, the switch 72 normally outputs the signal at terminal d, and outputs the signal at terminal C only when it is determined that a slope overload has occurred.

That is, when a decoded value in which a gradient overload has occurred is applied to the terminal d of the switch 72, the switch 72 outputs the decoded value of the next pixel applied to the terminal C, and the decoded value in which a gradient overload has occurred is applied to the switch 72. is replaced with the decoded value of the next pixel.

Next, a method for determining gradient overload will be explained.

Gradient overload occurs by limiting the D range of the quantization representative value as described above. Therefore, the value of a pixel where gradient overload occurs is always the decoded value at the end of the range. On the other hand, if a sample value corresponding to a large-amplitude impulse-like waveform as shown in FIG. 3(B) is input, for example, the amplitude is 0 at times other than time t3, and the amplitude is 255 at time t3. If the pixel is replaced with the decoded value of the next pixel, the information of the impulse (t=t3) will be lost. Therefore, the sign of the difference between the decoded value at the end of the range and the decoded values before and after it is detected, and only when the signs match, it is determined that a gradient load has occurred, and the decoded value of the next pixel is substituted.

Figures 3 (A) to (C) show the decoded values (indicated by ) and the decoded value after replacement (indicated by Δ).

The same figure (A) is a step waveform (steep edge), the same figure (B
) is the impulse waveform (thin vertical line part, etc.), same figure (C)
indicates input sample values and decoded values corresponding to a trapezoidal waveform (gentle edge). In both cases, the decoded value at time t3 is the edge end.

As is clear from FIG. 5A, in steep edge portions, a waveform equal to the input waveform is obtained by replacing the decoded values, and the influence of gradient overload is removed. on the other hand,
Regarding the impulse waveform, as shown in the same figure (B), the sign of the difference value before and after time t3 is positive before t3,
After , the replacement operation is not performed because the value is negative, and the information is not missing. Furthermore, in a gentle edge portion as shown in FIG. 2C, a decoded value replacement operation is performed although no large gradient overload occurs. In this case, a gentle edge becomes a steep edge, but
Considering the low sensitivity to differences in human visual characteristics, this will not be recognized as a deterioration in image quality, but rather the edges will become sharper, so the sharpness of the image will increase and the image quality will appear to have improved.

As is clear from the above description, in this embodiment, there is no deterioration in image quality due to gradient overload.

Furthermore, the sample values input to the encoding device are generally
Since the band is limited by the filter, Fig. 3 (B)
A large-amplitude impulse-like waveform like the one shown in Fig. 1 is almost never generated. In other words, since the amplitude of the waveform shown in FIG. 3(B) is reduced by the filter, the difference with the decoded value is reduced,
Considering that the sensitivity to differences in visual characteristics is low, deterioration is not detected visually.

Furthermore, when using two-dimensional prediction, three-dimensional prediction, or adaptive prediction as a prediction method, the prediction error becomes even smaller, and the probability that an impulse waveform will occur in two-dimensional space or three-dimensional space is almost negligible. It is 0, and there is no problem at all.

In this embodiment, the comparator 82 detects whether the difference value is positive or negative, but if the DPCM code is assigned so that the MSB (most significant bit) of the DPCM code indicates the sign of the difference value, the DPCM code It is possible to detect the positive or negative sign of the difference value using the MSB.

FIG. 4 is a block diagram showing an encoding device according to a second embodiment of the present invention. In the figure, 58 is an in-range determination circuit which will be described later, and those having the same functions as those in FIG. 1 are given the same numbers and symbols, and detailed explanation will be omitted.

In the first embodiment, the range end was determined using the encoded code and the quantizer was switched, but in this embodiment,
The quantizer is switched by the output of the in-range determination circuit 58 that uses the locally decoded value.

That is, the in-range determination circuit 58 uses the D-type flip-flop 5.
L
When l-+ is "L", when the local decoded value X1 is a decoded value corresponding to the range end of the range-limited difference value, Ll is set to "H", and otherwise, Ll is set to L". .

On the other hand, when L+-1 is "H", the local decoded value X1
When the decoded value corresponds to the end of the range of the range-limited difference value, that is, a value within the range, Ll is set to L'' and is 1; otherwise, Ll is set to 'H'''.

This judgment code Lr is a D-type flip-flop 56.
Apply to.

FIG. 5 shows a specific configuration of the in-range determination circuit 58 shown in FIG. 4.

When the predicted value x1 is applied, the range edge generation circuit 90 generates the maximum value X mx and the minimum @X- of the decoded value for the predicted value.
x to comparators 92 and 94, respectively. The maximum value Xmax and the minimum value X, 18 of the decoded value can be obtained by adding the quantized representative value at the end of the range of the D-range limited difference value to the predicted value. Therefore, by storing in advance the result of adding the predicted value and the range end of the difference value to the address indicated by the predicted value in the ROM, the range edge generating circuit 90
is obtained.

Comparator 92. At 94, the maximum value X□ of the applied decoded value, 8. The minimum value X□1o is compared with the local decoded value of 1, and data indicating the comparison result is applied to the switching determination circuit 96. The switching determination circuit 96 uses the determination result of the comparator 92.94 and the in-range determination signal L
i-! If is “L”, then X, = X,,,,
Alternatively, when X i = X mln, Ll is set to "H", and in other cases, Ll is set to "L". Also, Ll
-1 is “H”, when X min≦X1≦Xm, Ll is set to “L”, and in other cases, Ll is set to “H”.
shall be.

This in-range judgment signal is a D-type flip-flop 56.
is applied to.

That is, when the local decoded value or is at the end of the range of the decoded value by the DPCM fl quantizer, it is regarded as the edge part, and in order to suppress the influence of gradient overload, the PCM quantizer A is applied to the next pixel.
Select 34. On the other hand, when the decoded value by PCM quantizer A34 falls within the range of the decoded value by DPCM quantizer B36, it is regarded as a flat part and DPCM quantizer B36 is selected.

FIG. 6 is a block diagram showing an encoding device according to a third embodiment of the present invention. In the figure, 100 is a correction value generation circuit which will be described later, 102 is an adder, and 104 is a subtracter.

The edge portion normally has a low correlation with the previous pixel, but as described above, in the present invention, the quantizer switching signal is not sent, so the determination of the edge portion is delayed by one pixel. Therefore, the probability of appearance of a portion with a small difference value becomes somewhat high, and the PCM quantizer A34
The image quality will improve if you correct it.

The correction method of the PCM quantizer A34 will be specifically explained below.

The correction value generation circuit 100 converts the predicted value XI into a PCM quantizer A.
The quantization error Δx (ie, predicted value - quantized representative value) when quantized in step 34 is output to the subtracter 104 and the adder 102 . The subtracter 104 subtracts the correction value Δ× from the input sample value and applies it to the quantizer 34.

The quantizer 34 quantizes the corrected input value and converts it into PCM
The code YAI is sent to the terminal a of the switch 38 and the inverse quantizer 4.
2. The input sample value decoded by the inverse quantizer 42 is added with the correction value Δx by the adder 102, and is applied to the terminal a of the switch 50 as a decoded value X^1.

Now, assuming that the input value and the predicted value are equal (that is, the difference value is 0), the value applied to the quantizer A34 is smaller than the inverse quantizer 42 because the quantization error has been subtracted from the predicted value. The value applied to the adder becomes equal to the value applied to the quantizer 34, and by adding the quantization error in the adder 102, the input value is restored without the quantization error. Therefore, the difference When the value is 0, the quantization error is 0, so if the probability of the difference value appearing near 0 is high as described above, the quantization error is reduced, resulting in improved image quality.

Note that since the correction value is determined by the predicted value, no mismatch occurs between encoding and decoding.

In the above embodiments, the previous value prediction DPCM has been explained as an example, but the present invention is not limited to this, but can of course be applied to systems using prediction methods such as two-dimensional, three-dimensional, or adaptive prediction.

〔Effect of the invention〕

As described above, according to the encoding device of the present invention, since flat portions can be quantized with sufficiently fine quantization steps, generation of granular noise, false contours, etc. can be prevented. Also,
Since the sample values are quantized for the edge portion, it is possible to reduce the occurrence of gradient overload, and furthermore, it is possible to prevent error propagation in this portion.

Furthermore, according to the decoding device of the present invention, it is possible to decode in sufficiently fine steps in flat areas, without generating granular noise or false contours, and to reduce the occurrence of edge business in edge areas. This not only makes it possible to reduce the occurrence of gradient overload, but also to prevent error propagation.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a first embodiment of the encoding device of the present invention, and FIG. 2 is a configuration of an embodiment of the decoding device of the present invention corresponding to the encoding device of FIG. A block diagram showing the
), (B), CC') are diagrams showing how the decoding position is permuted in the decoding device of FIG. 2, and FIG. 4 is a block diagram showing the configuration of the second embodiment of the encoding device of the present invention. 5 is a diagram showing the configuration of the in-range determination circuit in FIG. 4, FIG. 6 is a block diagram showing the configuration of the third embodiment of the encoding device of the present invention, and FIG. 7 is a conventional code FIG. 2 is a block diagram showing an example of the configuration of a converting device. In the figure, 30 is a sample value input terminal, 32 is a subtracter, 34.36 is a quantizer, 38.50.70 is a changeover switch, 40 is an encoded code output terminal, 54.78 is a range end detection circuit, and 58 is a 62 and 64 are inverse quantizers, and 86 is a replacement judgment circuit. , 421 2nd drawing Quantity Shimiden □ Akebono □

Claims (4)

[Claims] (1) A first encoding means that quantizes a difference value between an input sample value and its predicted value, and a second encoding means that quantizes the input sample value.
encoding means, and selectively using the first and second encoding means to obtain an encoded code according to a change in the input sample value. (2) comprising a determining means for determining that the quantized representative value corresponding to the encoded code outputted by the first encoding means is at the end of the dynamic range of the difference value; An encoding device according to claim 1, characterized in that the first and second encoding means are selected. (3) determining means for determining whether the quantized representative value corresponding to the encoded code output by the second encoding means is within the dynamic range of the locally decoded value by the first encoding means; , the first and second according to the determination means.
The encoding device according to claim 1, wherein the encoding device selects the encoding means. (4) a first encoding means for quantizing a difference value between an input sample value and its predicted value; and a second encoding means for quantizing the input sample value.
A decoding device for decoding an encoded code obtained by selectively using an encoding means according to a change in an input sample value, the encoding means corresponding to the first and second encoding means, respectively. A decoding device comprising first and second decoding means, and selectively using the first and second decoding means to obtain a decoded value.