JPH02216917A - Coding/decoding method - Google Patents

Abstract

PURPOSE:To attain a coding method capable of highly effective transmission with no irregular error in simple constitution by transmitting an additional code including the estimating error information together with a coded remainder signal. CONSTITUTION:The maximum value SX and the minimum value SN of an estimating error being the difference between the input data on a block and its estimated value are obtained while using the block consisting of plural input data as a coding unit. When the prescribed value larger than (SX-SN) is defined as the divisor data OU, the remainder E obtained by dividing each input data of a block by the data OU is coded and transmitted together with an additional code including the estimating error information on the SX, SN and OU respectively. As a result, the reversible coding is made possible, therefore, an irregular error can be eliminated. Thus the decoding is also possible with no irregular error, and particularly the coding efficiency is further improved in a block which monotonously increases or decreases.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a highly efficient encoding method for encoding video signals, audio signals, etc. at a lower bit rate for digital transmission and recording. The present invention relates to a decoding method that performs inverse transformation.

Conventional Techniques Various methods have been proposed as highly efficient encoding methods. Among them, the differential pulse code modulation (hereinafter abbreviated as DPCM) method has characteristics such as a relatively simple circuit configuration and good quality of the reproduced signal, although the information compression efficiency is low.

Recently, as a new encoding method to replace the DPCM method, Ramamoorthy et al.
CM (hereinafter abbreviated as MPCM)) method is shown in the following documents [1]~
It is proposed in [3]. Furthermore, a new decoder for MPCM has been proposed by Hagiwara et al. in Reference [4].

(References) [1] Tay, Erickson and V, Ramamoorthy “Modulo-PCM: Anissource Coding Scheme, n-Inconf, Into-Conf,
Acoust, Speech, Signal Promoting, Icy Nisnisbee '791 Washington, D.C., 1979゜Bebe 419-422゜ (T, Ercson and V, Rasamo
orthy, “Modulo-PCM: Anew 5o
source coding scheme,'in Co
nf,,Int,Conf.

Acoust,,5peect++Signal pr
ocessing, ICASSP'79゜Washin
gton, DC+ 1979. pp419-422
．． ) [2] V, Ramamoorthy, ``Speech Coding Using Modulo-BCM With Side Information, ``Inconf, Rec 1. India, Conf, Acoust.

Speech, Signal Processing, Icy Niss. 81. Atlanta, Jeannie Ma, 19
81. Beebe, 832-835° (V, Rama
moorthy,”5peech coding
using Modulo-PCM with 5i
de information,”in Conf,
Rec,, 1nt.

Conf, Acoust, 5peech, Signa
l Processing, ICASSP'81+A
t1anta+G^, Mar, 1981+pp, 83
2-835. ) (3) V, Ramamoorthy “A Nobel Speech Coder for Medium and High Bit Late Applications Using Modsieropcm Principles,” Ieeeeee Trance, Acoust, Speech 1
Signal Processing, Pol, Ninis Sp3
3. PP, 356-368°NIB R, 1
985゜(V, Ramamoorthy,'＾ Novel
5peech Coder for Medium
and High Bit Rate Applica
tions Using Modulo-PCM Pr1
nciples+'1EEE Trans, Acou
st++5peech, Signal Process
ing+vol1. ASSP331pp, 356368,
April, 1985) [4] Hagiwara, Nakaso, "New decoder for Modulo-PCM", Information Theory and Its Applications Research Group 8th Symposium+ (Nara+ j apan, Dec.
, 5-7.1985+pp517-522) The MPCM method has almost the same circuit configuration as the DPCM method, and is the same as the DPCM in terms of the presence of granular noise, but has the advantage of not having overload noise. There is.

First, the MPCM method will be briefly explained.

The basic system of the MPCM method is shown in FIG.

The MPCM encoding device consists of a remainder calculator 7 (11) and a quantizer 702.

Here, the sample input signal is xK.

The remainder calculator has sawtooth input/output characteristics with an amplitude d, as shown in the figure. and its output (
(xK)) can be expressed as ((XK))−x*d·((XK)). Here, ((XK)) is an arbitrary constant that satisfies the following formula (A).

(-d/2)≦((x,))<(d/2)・・・・・・
-(A) The output ((X, )) of the remainder calculator is quantized into L bits to become a transmission signal yK, which is transmitted to the decoding device.

In this way, in the MPCM encoding device, the respective waveform diagrams are shown in FIG. 10, and as shown in FIG.
This is P
In the CM system, this is equivalent to transmitting only the lower bits.

In the decoding device, as shown in Fig. 9, the first-order predictor calculates
), and predict the value xK of the sample point from (I! The upper bit component d·[[χE]] which has not been transmitted is determined, and this is added to the transmission signal) lx to reproduce the input signal of the encoding device.

Problems to be Solved by the Invention Next, irregular errors, which are a major problem in the MPCM system, will be described.

If the absolute value of the input signal difference value ΔX x =X KX *−1 becomes larger than d/2, the decoder will not correctly predict the upper bit component d·[[xK]]. FIG. 12 shows a waveform diagram in that case. Therefore, the decoded output signal xK has many points where the waveform changes sharply, as shown in FIG. 12(C). This is an error unique to the MPCM method called an irregular error, and the challenge of the MPCM method is how to reduce this error.

Therefore, Mr. Ramamoorthy et al. proposed a method of suppressing the occurrence of irregular errors as much as possible, which is shown below (References (2) and (3)).

Method 1: In this method, a decoding device is installed in the encoding device, and when an irregular error occurs in this decoding device, it is detected and information for correcting the irregular error is transmitted as side information on a separate channel. be.

Method 2: When the prediction error is large, irregular errors occur frequently, so the input data is grouped into blocks every N samples, the maximum value EaN of the absolute value of the prediction error is determined for each block, and the modulo of remainder operator
controls the width of the quantizer and the quantization step of the quantizer. In reality, the maximum value E,,I is quantized and multiplied by a predetermined coefficient, which is called the modul.

, and this dlIN is transmitted to the decoding device as side information.

However, the above two methods have the following problems.

In the case of method 1, since it is uncertain which sample of the encoded input data an irregular error occurs, it is necessary to send correction information for each input sample in order to completely eliminate the irregular error. This is equivalent to increasing the number of quantization bits, and the encoding efficiency decreases. When the number of quantization bits of the quantizer is large, irregular errors occur less frequently, so the bit rate of the channel that transmits side information can be lower than that when transmitting information that corrects each sample. It is not possible to eliminate irregular errors.

In the case of method 2, the frequency of occurrence of irregular errors can be lowered by optimizing various parameters, but it does not fundamentally eliminate irregular errors due to the influence of quantization errors caused by the quantizer, so irregular errors cannot be completely eliminated.

Furthermore, Hagiwara et al. have proposed a decoding device that corrects irregular errors by providing a new determination circuit on the decoding device side without sending side information. However, since it is not possible to detect all irregular errors, it is not possible to completely eliminate irregular errors (Reference (4)).
.

Furthermore, when the input of the MPCM method is an image signal, irregular errors cause a visually noticeable deterioration in image quality, so application of the MPCM method to image signals has not been considered.

Means for Solving the Problems The present invention uses a block composed of a plurality of input data as a unit of encoding, and calculates a maximum value SX and a minimum value SN of prediction errors, which are the differences between the input data and its predicted value in the block. When a predetermined value larger than (SX-SN) is set as the divisor data OU, the remainder E obtained by dividing each input data of the block by the divisor data OU is encoded, and the remainder E obtained by dividing each input data of the block by the divisor data OU is encoded.

This encoding method is characterized in that an additional code containing prediction error information regarding SN and OU and the encoded E are transmitted together.

The present invention also provides a code in which, in the above encoding method, a signal obtained by converting the original input signal so that the dynamic range of the prediction error of the original input signal is equal to or less than a predetermined amount for each predetermined number of blocks is used as the input signal. It is a method of conversion.

Further, the present invention uses a block composed of a plurality of input data as a unit of encoding, and provides a maximum value SX and a minimum value S of prediction errors, which are the differences between input data and its predicted value in the block.
N, and when a predetermined value larger than (SX-SN) is set as the divisor data OU, the remainder E obtained by dividing each input data of the block by the divisor data OU is encoded, and the SX, A signal encoded by a coding method that transmits both an additional code containing prediction error information regarding SN and OU and the encoded E is input, and the divisor data OU is obtained from the transmitted additional code and transmitted. A prediction error obtained by adding the divisor data OU multiplied by an integer N to the decoded remainder to obtain decoded data, and the prediction error of the difference between the decoded data and its predicted value is obtained from the transmitted additional code. The decoding method is characterized in that the integer N is determined so as to satisfy a range.

Further, the present invention is a decoding method characterized in that, in the above-described decoding method, when the decoded data is converted from the original input signal, the original input signal is obtained by performing inverse transformation.

Function: The present invention can perform reversible encoding with the above-mentioned configuration, so irregular errors can be completely eliminated. Furthermore, when constant rate encoding is performed, reversible encoding is performed after the original input signal is transformed so that the dynamic range of the prediction error becomes less than a predetermined amount, so that decoding can be performed without irregular errors. Furthermore, since the divisor data is determined based on the dynamic range of the prediction error, encoding efficiency can be further improved especially in monotonically increasing or monotonically decreasing blocks.

Embodiments Before giving detailed explanations of embodiments of the present invention, the principle thereof will be explained.

Multiple input data with close sample positions (DI, D2゜...
..., Dk) as a unit of encoding, if you take out the input data in the block one by one and calculate the difference, the difference value will be small due to the correlation of the signals, and the In most blocks, the dynamic range of the difference value (the difference between the maximum value and the minimum value) is also smaller than the dynamic range of the input data. This difference value is the prediction error when predicting the previous value, so the prediction error S
The present invention is based on the property that the dynamic range of this prediction error Sl is smaller than the dynamic range of input data in most blocks, and attempts to perform more efficient encoding than conventional methods. It is.

When the maximum value, minimum value, and dynamic range of the prediction error Si in the block are SX, SN, and SDR, the following formula % formula % (1) holds true: 0th order formula OU>SDR... 2) A predetermined value that satisfies the following and is uniquely determined by the dynamic range SDR is defined as the divisor data OU.

Input data Di in block (i-1, 2,...
When the quotient and remainder obtained by dividing .

Transforming the formula (3), the following formula Ei =D I -0U-N i...
-(4) is obtained. The remainder Ei is OU from the input data Di.
・It can be considered that the Ni offset has been removed.

Let the offset be expressed as Fi. That is, the offset Fi is defined by the following equation (5).

The dynamic range SDR of the prediction error is mostly blocks that are smaller than the dynamic range of the input data, so in Equation (2), if the divisor data OU is set to a small value, the divisor data OU will be smaller than the dynamic range of the input data. Therefore, if the input data Di can be restored from the residual data Ei, a more efficient code can be obtained by encoding and transmitting the residual data El instead of the input data Di. The encoding method of the present invention, which can perform 0 conversion, realizes this.

Hereinafter, input data DI is obtained from the residual data El. That is, the decoding method will be explained.

In order to return the surplus data Ei to the input data Di, it is necessary to reproduce the offset value Fis=0υ·NI.

If the predicted value for input data Di is Pi, the maximum value and minimum value of the prediction error are SX.

Since it is SN, the following formula % formula % (6) holds true, and the input data D+ can be expressed by formula (3), so this is substituted into formula (6) and the following formula Pi + SN - Ei ≦ OU・Ni ≦Pi+5X-Ei...
...(7) is obtained. The predicted value Pi (+=2.3...) can be obtained sequentially from the input data obtained by decoding (7), and in addition to SX, SN, 0U (7), Pl or Fl
(-0U-Nl) or N1 may be known.

Since previous value prediction is performed, prediction (!PI can be data in the adjacent block that has already been decoded,
It is not necessarily necessary to transmit, therefore, the encoding device transmits information regarding prediction error SX and SN for each block along with the residual data Ei.

If OU is transmitted, off-sera) F 1=O according to equation (7)
U-Ni can be uniquely determined, data Di is obtained by adding the remainder data Ei, and decoding is completed. Furthermore, by adding PI, Fl, or Nl to each block and transmitting it, each block can be encoded and decoded independently because the input data of other blocks is not used, making it highly resistant to transmission errors. can.

The offset Fi is uniquely determined by the definition of OU (
This is because the difference between Pi+5n-Ei) and (Pl+5X-Ei) is smaller than OU. Therefore, 0, U-Ni2(
P i+SN-E i)>OU・(Ni-1)...
...0H-Nl that satisfies (8) is also an offset Fl, and 0U-
Ni5(Pl+5X-El)<OU・(Ni+1)
......(9) 0U-Ni that satisfies the following is also an offset Fi.

That is, the same offset Fi can be obtained using any of equations (7), (8), and (9).

Since the remainder data Ei is the remainder obtained by dividing the input data D+ by the divisor data OU, it is greater than or equal to O and less than the divisor data OU. Therefore, the word length of the remainder data Ei is determined by the size of the divisor data oU, and is variable in length for each block.

The number of points required to represent the OU is 1υg20U, which decreases as the OU becomes smaller. However, since the word length of this residual data is an integer, it satisfies the following formula (if 2 ” or less). Even if the divisor data OU is used, the word length remains unchanged and the encoding efficiency is not improved.

Rather, by using the 2H as the divisor data OU, the OU
Various advantages can be obtained, such as a smaller number of bits required to represent the offset, and a simpler circuit (for example, a divider) for determining the offset.

This point will be discussed later.

When determining the offset value Fi necessary for obtaining data D+ from the residual data Ei in the decoding device, SX,
I mentioned that SN, 0υ are necessary, but SX, SN, O
Instead of U, it is possible to transmit predetermined values SXX, SNN, and OU that satisfy the following equation together with the residual data El, and use these to similarly determine the offset 4flFi required for decoding. This is because the following formula % formula % (However, when the left equal sign is SNN-SN, the right equal sign is 5
This holds true when xx-sx. ) holds true.

Since SX, SN, SDR, OU, SXX, and SNN are information regarding prediction errors, they will be collectively referred to as prediction error information. SX, SN, OU or SXX.

Among the three data combinations of SNN and OU, one can be determined from the remaining two data, so any two of the three data combinations can be transmitted directly or indirectly as prediction error information. , and since the divisor data OU is uniquely determined from the dynamic range SDR of the prediction error, SDR (-3X-S
N) may be transmitted as one of the prediction error information.

In the above explanation, previous value prediction was used for prediction, but more advanced prediction may be used.

Further, according to the above-mentioned encoding method, the prediction error information (SX, SN, OU) and the residual data obtained by encoding are each an integral multiple of the divisor data OU than all the input data Di in the block. It is equal to the prediction error information obtained by encoding the data obtained by removing the constant value and the residual data. This is because even if a constant value is removed from all input data, the prediction error remains the same, and if the constant value is an integral multiple of the divisor data OU, the remainder data that is the remainder obtained by dividing by the OU will also be the same. Therefore, the following decoding method is also possible. In other words, all the residual data Ei (1-1, 2,...
..., k) and prediction error information (for example, SX and OU), the initial value Fl (or PI, or Nl)
Even if the input of is not completed, it is possible to proceed with decoding as Fl-0, and by adding the initial value Fl obtained thereafter to (Di-Fl) obtained as a result, the decoded data Di
That is.

FIG. 1 is a block diagram showing the configuration of a highly efficient encoding device in a first embodiment of the present invention.

In FIG. 1, 1 (11 is an input terminal for input data that is a sampled and quantized video signal, 102 is an input terminal for input data that is a sampled and quantized video signal, and 102 is a screen that the video signal represents is divided into small blocks, and input data Di (i=1. 2. 103 is a block divider that outputs .
4 is a register that holds the first data D1 of each block output from the delay device 103; 1 is a register that receives the output from the block divider 102, and outputs the previous input data as the predicted value Pi; Predictor, 10
6 subtracts the predicted value Pi from the output of the block divider 102 to obtain the prediction error Si (1=2.3...
16), 1 (17 is a subtractor 1 for each block)
A maximum/minimum detector calculates the maximum value SX and minimum value SN of the prediction error Si from 06, holds it in an internal register for each block, and outputs it. 108 subtracts the minimum value SN from the maximum value SX and makes a prediction. A subtracter 109 outputs the dynamic range SDR of the error, and 109 is the dynamic range SD.
A divisor data generator 110 which receives R as input and outputs divisor data OU divides input data Di from the delay unit 103 by the divisor data OU to generate remainder data Bi (+-2
, 3, ..., 16), 111
112 is a remainder encoder which encodes the remainder data Ei according to the size of the divisor data OU to obtain encoded data Ci; 112 is an output terminal for the encoded data Ci; 113
114 is an output terminal of the maximum 41SX of the prediction errors, and 115 is an output terminal of the first data DI in the block.

The operation of the encoding device of this embodiment configured as described above will be described below.

Input data (word length 8 bits with amplitude level .theta..about.255), which is a sampled and quantized video signal, is input to the block divider 102 from terminal 1 (11).

The block division circuit 102 has an internal buffer memory, and simultaneously writes input data into the buffer memory in order, and at the same time writes input data D1 (i=1.2...) in block units.
..., 16) is read out. Input data Di outputted first from each block is held by the register 104 and outputted from the terminal 115 as one of the additional codes.

The input data input to the predictor 1 (15) is output as a predicted value pt and is supplied to the subtracter 106.
In step 6, the predicted value Pi is subtracted from the input data Di to obtain a prediction error St. Maximum minimum detector 1 (17
, the prediction error Si is input, and the maximum value SX and minimum value SN of the prediction error St within the block are output. The minimum value SN is subtracted from the maximum value SX by a subtracter 108 to obtain the dynamic range SDR of the prediction error. The dynamic range SDR is input to the divisor data generator 109, and divisor data OU satisfying the following formula % is output. Delay device 103
The input data Di whose timing has been adjusted by is divided by the divisor data 0υ in the remainder calculation unit 110 and outputted as remainder data Ei (i=2.3° . . . , 16). Since OU is 2H, all of the residual data E, which is less than 10, can be expressed with a word length of M bits without deterioration.

Further, since the remainder data E1 is never larger than the input data Di, its word length is 8 or less. The remainder data Ei input to the remainder encoder 111 has the number of bits J determined by the size of the divisor data OU (where M<9
When J-M, when M-9, J=8), the data is encoded as variable-length encoded data Ci and output from the terminal 112.

, since non-linear processing such as truncating lower bits is not performed on the residual data Ei during encoding, no deterioration of the residual data Ei occurs due to encoding, that is, reversible encoding is performed. Therefore, encoded data CI
The residual data EI° obtained by decoding is inferior to the original residual data E1.

The divisor data OU, which is prediction error information, and the maximum value SX of the prediction error are output from terminals 113 and 114, respectively, by the additional code 1.
output as one.

FIG. 2 shows input data Di when the block configuration is 16 pixels in the horizontal direction and 1 line in the vertical direction.

The 16 input data in this block are D
112.1 (15,113°121.13
6,160,175,2 (11,2 (17°208.1
99,193,176.169,178°190. In the case of such a one-dimensional block configuration, the block divider 102 does not require an internal buffer memory, and since the predictor 1 (15) only needs one register to hold the previous input data, the encoder The configuration is simpler.

The order in which data is extracted from two-dimensional blocks and the prediction method will be described later.

The maximum and minimum values of the maximum possible prediction error SX are 255 and -255, respectively, and the minimum and minimum prediction errors (a S N are also 255 and -2, respectively.
55, the word length of the maximum value SX and minimum value SN of the prediction error requires 9 bits.

Maximum value of dynamic range SDR of prediction error.

Since the minimum value is 510.0, the word length of the dynamic range SDR needs to be 9 bits.

Therefore, the word length of the divisor data OU generally needs to be 9 bits, but in this embodiment, the divisor data OU is set to be 2 to the M power that satisfies the formula QlIl, so a smaller number of bits is required. It is possible to represent That is, instead of directly transmitting the divisor data OU, it is sufficient to transmit the exponent M in a power-of-two representation. The index M is 0, 1, 2.・・・・・・Since there are 10 ways of 9, the word length only needs to be 4 bits, that is, the divisor data Oυ is 4 bits.
It is transmitted using the bit code M.

Divisor data OU is the dynamic range of prediction error SDR
Since the divisor data generator 10 is uniquely determined by
9 can be configured with a 9-bit input, 4-bit output ROM (read only memory). However, since the divisor data Oυ is a power of 2, it is possible to configure the divisor data generator 109 with a simple logic circuit that detects the number of consecutive O's from the most significant bit in the data representing the dynamic range SDR. .

The remainder arithmetic unit 110, which obtains remainder data Ei by dividing input data by the divisor data OU, can also generally be configured with a divider or a ROM. However, in this embodiment, the divisor OU is a power of 2, so the remainder Et after dividing by the OU is the lower J bits of the input data Di (J-M when M<9, J when M-9). The remainder arithmetic unit 110 can be realized by a simple logic circuit that extracts only the bits (-8) and sets the other bits to 0.

a remainder encoder 11 that encodes the remainder data Ei into the number of bits J determined by the size of the divisor data OU;
1 extracts only the lower J pit of the residual data Ei,
This can be realized with a simple circuit that does not output unnecessary upper bits. There are various possible output formats for the encoded data Ci, but in this embodiment, it is assumed that the encoded data Ci is output in the form of serial data.

Although omitted in the figure, additional codes (DI, SX, OU), which are output signals of the encoding device, and encoded data (C2
, C3, . . . , C16) are stored in a buffer memory, subjected to error correction encoding processing, etc., and then output to a transmission line in the form of a serial data signal, for example.

In the encoding device of this embodiment, the total number of bits of input data Di (i = 1.2..., 16) per l block is 12B (=8 x 16) pits, and the output per l block is The total number of bits CD is an additional code of fixed word length (
21 (-8+9+4) bits of DI, SX, OU) and variable length encoded data (C2, C3,..., C1
6) with 15·J bits, and can be expressed by the following formula: %:1. The word length J of the encoded data Ci is 0 or more and 8
As shown below, in most blocks, the prediction error is small due to the correlation of the input signals, and the word length J is 7 or less, so the following formula % formula % is established, and the The average number of bits can be reduced. Note that when the word length J is 8.9, the encoding efficiency deteriorates, so it is better to transmit the input data as is. Fig. 2 shows an example of each data in a one-dimensional block. According to the figure, the divisor data OU is 64.
Therefore, the word length J of the encoded data Ci is 6. Therefore, the total output focus number CD of the encoding device in this block is 111 focus points (<128) due to the number 2.
The average number of bits per sample is reduced to approximately 6.9 bits.

FIG. 3 is a block diagram showing the configuration of a decoding apparatus according to the first embodiment of the present invention, and this decoding apparatus performs inverse transformation of the encoding apparatus shown in FIG.

In FIG. 3, 2 (11 is the input terminal of encoded data Ci, 202 is the input terminal of divisor data OU, 203 is the input terminal of the maximum prediction error SX, and 204 is the input terminal of the first data D1 in the block. , 2 (15 is a unit that performs the inverse transformation of the remainder encoder 111 in the highly efficient encoding device shown in FIG. 1, and decodes the encoded data Ci to generate remainder data Ei.

A remainder decoder 206 obtains the maximum value S of the prediction error.
Adder 2 with x as one input (17 is the adder 2
208 is an offset regenerator that receives the output of the subtracter 2 (17 and the divisor data OU and outputs the offset Fi'); 209 is the offset regenerator that receives the output of the subtracter 2 (17) and the divisor data OU, and outputs the offset Fi';'' and the offset Fi' to obtain decoded data Di°; 210 is a switch that inputs the output of the offset adder 209 and the data D1 from the terminal 204 and selects one of them; 211 is predictor 1 in FIG.
(Same as 15, the predictor inputs the decoded data Di and supplies the predicted value Pi to the other input of the adder 206; 212 is the decoded data Di from the switch 210;
A block decomposer 213 which receives the signal .degree. and performs the opposite process to that of the block dividing circuit 102 in FIG. 1 is an output terminal for decoded data.

Note that the additional code input precedes the encoded data input, and l
It is assumed that the block period is maintained.

The operation of the decoding device configured as above will be explained.

The input data D1 from the terminal 204 is outputted as decoded data DI' via the switch 210 and supplied to the block decomposer 212 and the predictor 211. The data DI' becomes a predicted value P2' necessary for the predictor 211 to obtain decoded data from the initial residual data E2'.

After outputting the input data DI, the switch 210 selects the input from the offset adder 209 and outputs the decoded data D2.
' , D3°, ..., D16° are output.

Encoded data CI from terminal 2 (11 (i=2゜3
, . . . , 16) is converted into remainder data EBI by the remainder decoder 2 (15) based on the divisor data OU from the terminal 202. The maximum value SX of the prediction error from the terminal 203 and the prediction The predicted value Pbi from the adder 211 is added to the adder 206.
The remainder data Ei is added from the addition result.
° is subtracted by subtractor 2 (17
B) is obtained. The result of the subtraction and the divisor data OU are input to the offset regenerator 208, and an offset Fi'=OU·Ni that satisfies the following equation (%) is obtained.

In the offset adder 209, the offset F i
' and the residual data Ei'' are added to become decoded data Di (i=2.3......, 16),
Supplied to switch 210.

At the output of the switch 210, the decoded data D i '(i = l, 2...5
16) All obtained predictor 211 and block decomposer 2
12. The decoded data Di'' input to the block decomposer 212 is extracted in the same order as the data obtained by sampling and quantizing the original video signal, and is extracted from the terminal 213.
It is output from

The encoded data Ci input in the form of serial data from the terminal 2 (11) has a variable word length for each block, so in order to correctly extract the encoded data Ci, it is necessary to know the word length J. be.

The word length J of the encoded data Ci is determined by the index M representing the size of the divisor data OU.
can be determined (however, when M<9, J-M, M=
9 (J=8), encoded data Ci can be extracted accurately, and residual data Ei' can be obtained by adding O to its upper bits. Therefore, remainder decoder 2
(15 can be realized with a simple circuit such as a shift register for serial-to-parallel conversion, a register for holding encoded data, a counter for counting the number of points in focus, and a gate for adding O to the upper bit.Remainder encoder 111 in FIG. There is no deterioration due to encoding, and the remainder decoder 2 (15) performs inverse transformation of the remainder encoder 111, so the remainder data Ebi matches the remainder data Ei.

By dividing (Pbi+SXE+'), which is the output of the subtracter 2 (17), by the divisor data OU to obtain a remainder, and subtracting the remainder from the (P i' +5X-E i'), equation (9) is obtained. ) is obtained.However, in this embodiment, the divisor data OU is set to M of 2.
Since it is a power of
can be obtained. Therefore, the offset regenerator 208 can be realized with a simple logic circuit.

The residual data Ej and the predicted value Pi' in the decoding device are the residual data Ej and the predicted value P in the encoding device in FIG.
Since it is equal to i, the offset Fi is also equal to the offset Fi, and the decoded data Di' matches the input data DI.

As described above, according to this embodiment, there is no quantizer like MPCM, the residual data is encoded and transmitted as is, and the variation range of prediction error, divisor data OU (MPCM system modulo0) is set larger than the range of change (dynamic range) of the prediction error, and furthermore, this divisor data OU and information on the range of change of the prediction error are transmitted as additional codes, so irregular errors and quantization errors are avoided. In other words, reversible encoding and decoding can be realized, and the circuit configuration thereof is extremely simple.

Furthermore, in the conventional example, divisor data (mo
dulo width) was fixed or the maximum absolute value of the prediction error, but in the present invention it is determined by the variation range (dynamic range) of the prediction error, so the divisor in encoding can be smaller than in the conventional example. Therefore, more efficient encoding can be performed than in the conventional example especially in blocks where input data monotonically increases or decreases. When an image signal is input, a rough image can be reproduced using only a part of the additional code (the data of the first pixel of the block), making it possible to perform encoding suitable for image retrieval.

Specific position within the block! An encoding method may also be considered in which the average value DA of input data within a block is transmitted as one of the additional codes instead of the input data.

In this case, the decoding method is to first decode the off-foot initial value F1-0, and the obtained data (Di-Fl)
Find the average value DB. Transmitted average (tI D A
By subtracting the average value DB, the initial value F1 of the offset can be obtained. Decoded data DI is obtained by adding the obtained initial value Fl to the data (Di-Fl). If all of the residual data Ei are transmitted, even if one error occurs in the residual data, it can be corrected by using the average value DA. Furthermore, when the input is an image signal, the outline of the image signal can be represented using only the average value DA, and encoding suitable for searching can be achieved.

Next, a case where an error occurs in encoded data on a transmission path will be explained.

Even if a code error occurs on the transmission path, the error does not necessarily propagate to the decoding of the next sample as in the DPCM method. This is because there is a range of judgment determined by the maximum and minimum values of the prediction error.

However, an error may occur in determining the offset Fl, and the error in the offset Fi is propagated to subsequent decoded data.However, when obtaining the predicted value Pi, data in other blocks is not used. In other words, since each block is encoded independently, errors will not propagate to neighboring blocks, and the difference between correct decoded data and incorrectly decoded data due to error propagation is an integer of the divisor data OU. Therefore, it is possible to accurately correct the error by determining the correlation between the decoded data in which the error has been propagated and the correct decoded data next to it.

We will explain how to correct it. The position where the error occurs can be detected by using an error correction code.For example, suppose that a code error occurs in the ninth residual data shown in FIG. 2(C) and the offset Fi is determined incorrectly.

As a result, the subsequent 10th to 16th data will also be decoded incorrectly, so the 17th data (not shown in Figure 2), which is the first data of the next block where the error will not propagate, is used as the starting point. Decoding is performed while predicting in the opposite direction to that used during encoding.

As long as the difference between the 17th data and the 16th data does not exceed the dynamic range of the prediction error, this will result in 16
It is possible to correctly decode up to the 10th, 15th, . . . , 10th input data. However, since the prediction direction is different from that during encoding, the sign of the prediction error is inverted, and the maximum and minimum values used for decoding are the inverted signs of the minimum and maximum values transmitted by the additional code, respectively. There is a need. The 9th input data can be modified using the average value of the 8th and 10th decoded data.

Also, the prediction error obtained by performing prediction across block boundaries using decoded data (in the example in Figure 2, the 16th and 1st
It can be determined that there is a high possibility that a transmission error has occurred in an area where the value (corresponding to the difference in the seventh data) is outside the range of the prediction error determined from the additional code. If it is determined that it is a transmission error, the out-of-range area can be predicted and decoded from a direction different from the direction used during decoding, starting from the decoded data in the adjacent block, as in the above method.

In the case of FIG. 2, the block configuration is one-dimensional, but if it is two-dimensional or more like an image signal, it is possible to correct it using decoded data in a plurality of surrounding blocks. If a transmission error occurs in the additional code representing prediction error information, a correction method using the additional code of an adjacent block may be considered. Furthermore, in the event of a transmission error, an encoding method may be considered in which it is determined at the time of encoding which adjacent block is appropriate for use in correction, and this information is transmitted as one of the additional codes. It becomes possible to do this.

The first embodiment described above is variable-length encoding in which the number of encoded bits changes for each block, and is more suitable for applications that do not require real-time transmission, such as file recording of still images, short audio, etc. There is.

Next, the principle of the encoding method of the present invention, which is suitable for a normal transmission line with a constant bit rate, will be briefly explained.

One way to keep the bit rate constant is to keep the number of encoded bits constant for each block. Keeping the number of encoded bits constant means that the dynamic range of prediction error is kept below a certain amount. In the present invention, the input signal is transformed to reduce the dynamic range of its prediction error to a predetermined amount or less to obtain a transformed input signal, and the reversible method of the present invention is applied to the obtained transformed input signal. It performs encoding. Further, the decoding according to the present invention described above is performed on the encoded output to reproduce the transformed input signal, and the input signal is reproduced by performing the reverse transformation to that during encoding.

The simplest method for obtaining the transformed input signal is to multiply the input signal by a predetermined number for each block. By multiplying the input signal by a predetermined number smaller than 1, a converted input signal with a reduced dynamic range is obtained,
The dynamic range of the prediction error of the converted input signal can be reduced by approximately the predetermined number of times the dynamic range of the prediction error of the input signal. This is because the process of obtaining a prediction error from an input signal is linear.

FIG. 4 shows a block diagram of an encoding device and a decoding device according to a second embodiment of the present invention. This encoding device maintains a constant number of encoded bits for each block.

On the right side of Fig. 4, 4 (11 is an input terminal for input data which is a sampled and quantized video signal (word length 8 bits, 0 to 255 levels), and 402 is an input terminal with a constant word length of encoded data for each block. 403 is a decoding device that performs inverse transformation of the encoding device 402, and 404 is an output terminal for the decoded input signal.

In the encoding device 402, 102 is a block divider (same as the one in FIG. 1), 4 (15,406°4 (17.
408 is a coefficient of 1/2.1/4°1/8.1/1 respectively
Coefficient multiplier that multiplies input signal by 6, 409.410,411
, 412 is an encoding device that has an internal configuration from which the block divider 102 is removed in FIG. a detector that receives divisor data OU as input and outputs conversion information as one of the additional code outputs;
414 the encoding devices 409 to 412 and the coefficient unit 40;
The output of B is input, and if there is an encoded data whose word length is 4 bits from among the outputs of the encoders 409 to 412 under the control of the conversion information, that output is selected and output; otherwise, the coefficient is This is a switch for selectively outputting the output of the device 40B.

In the decoding device 403, 212 is a block decomposer (same as the one in FIG. 3), 415 is a decoding device having an internal configuration from which the block decomposer 212 in FIG. 3 is removed, and 416 is the output of the decoding device 415. and a switch which inputs encoded data and outputs one of them under the control of the transmitted conversion information; 417 is a switch which selectively outputs a predetermined coefficient under the control of the transmitted conversion information;
A multiplier 418 multiplies the converted signal reproduced from the switch 416 by a predetermined coefficient from the switch 417 to reproduce the input signal.

The operation of the encoding device of this embodiment configured as described above will be described below.

In the encoding device 402, the input signal from the terminal 4 (11) becomes a signal for each block by block division 8102.
8, the predetermined coefficient 1/2°1/4.1/8.1/16
are multiplied and rounded to the nearest whole number to create converted input signals (including input signals) with word lengths of 8, 7, 6, and 5.4 bits, and each is reversibly encoded by encoding devices 409 to 412.

Detector 413 detects an encoding device whose encoded data has a word length of 4 bits. The switch 414 is the detector 41
The output of the encoding device whose encoded data has a word length of 4 bits is selected and output under the control of the output of step 3, but it is not possible to detect an encoding device whose encoded data has a word length of 4 bits. If not, the output of the coefficient unit 408 is selected and output as encoded data. Since it is necessary to inform the decoding device of information (3 bits) indicating which converted input signal was encoded and transmitted, this information is included in the additional code and transmitted as conversion information. Note that the encoding device 40
9 assumes that the lower limit value of the encoded data word length is set to 4 bits. Therefore, the amount of information to be transmitted is 80 bits in total, with the additional code being 3 bits of conversion information, 8 bits of block head data (worst value), the maximum prediction error being 9 bits (worst value), and the encoded data being 4 x 15 bits. ,lN
The average per bit is 5 bits.

In the decoding device 403, a decoding device 415 decodes the transmitted encoded data and decodes the converted input signal in the encoding device without error. The decoded converted input signal becomes an input to a multiplier 418 via a switch 416. The switch 417 selects and outputs an inverse transform coefficient for converting the decoded transform input signal into an input signal under the control of the transmitted transform information.The 0 multiplier 41B converts the inverse transform coefficient from the switch 417 into the transform input signal. Convert to input signal by multiplying by The reproduced input signal for each block is decomposed into blocks by a block decomposer 212, and an input signal having the same data arrangement as the input signal of the block divider 102 is sent to the terminal 4.
Output from 04.

Note that if the transmitted encoded data is detected to be the output of the coefficient unit 40B based on the transmitted transformation information, the encoded data is the transformation input signal itself, so it is directly switched without going through the decoding device 415. Multiplier 41 via 416
This becomes the input of B.

When the input signal is encoded as described above and cannot be encoded with a predetermined encoded data word length, the input signal is multiplied by a predetermined coefficient to create a transformed input signal with a small dynamic range of prediction error, and this is encoded. By doing so, the encoded data word length can be kept constant. Since the encoding of the converted signal is reversible using the same encoding method as in the first embodiment, no irregular errors occur.If the dynamic range of the prediction error of the input signal is large, the input signal is multiplied by a predetermined coefficient and the converted signal is input. Since it is converted into a signal and then encoded, the quantization becomes coarse. Places where the dynamic range of prediction errors is large are places where the input signal changes rapidly. In areas where the signal changes rapidly, even if the quantization is coarsened, no signal deterioration will be felt, so it can be used for practical encoding that matches visual characteristics.

Also, for ease of explanation, there are four encoding devices 409 to 41.
Although most of the circuits inside the encoding device can be shared, the circuit size does not increase significantly compared to the first embodiment.

In the second embodiment, the transformation coefficients for converting the input signal into the transformation input signal are fixed and there are only a few types, so the dynamic range of the encoded data may not be utilized effectively. this! An embodiment of the present invention that solves the problem will be shown below.

FIG. 5 shows a block configuration diagram of an encoding device according to a third embodiment of the present invention.

In FIG. 5, 5 (11 is an input terminal for input data which is a sampled and quantized video signal, 102 is a block vI
(same as 102 in FIG. 1), 502 is a conversion circuit that generates conversion information and conversion coefficients, multiplies the input signal by the conversion coefficient, and outputs a converted input signal; 503 is an encoding device;
04 is an output terminal of the transformation information (inverse transformation coefficient 1/a);
5 (15 is the first data d in the block of the conversion input signal,
506 is the output terminal of the maximum prediction error value SX for the converted input data, 5 (17 is the encoded data ci
This is the output terminal of

In the conversion circuit 502, 508 is a delay device, 509 is a predictor, 510 is a subtracter, and 511 is a maximum/minimum detection circuit, which corresponds to the delay device 103, predictor 1 (15,
subtractor 106, maximum/minimum detection circuit 1 (equal to 17, 5
12 is conversion information (
A conversion information generation circuit 513 outputs a conversion coefficient a, an inverse conversion coefficient 1/a), and a multiplier 513 multiplies the input signal by the conversion coefficient and outputs a converted input signal.

Components in the encoding device 503 that are the same as those in FIG. 1 are given the same numbers. 103 is a delay device, 104 is a register, 1 (15 is a predictor, 106 is a subtracter, 110 is a remainder operator, III is a remainder encoder, 514 is a subtracter 10)
This is a maximum value detection circuit that detects the maximum value of the prediction error for each block from 6 and stores it in an internal register for each block and outputs it.

The operation of the encoding device of this embodiment configured as described above will be described below.

Input data, which is a sampled and quantized video signal, is sent to terminal 5.
11, and is converted into a signal for each block by the block divider 102.The input signal for each block is multiplied by a conversion coefficient a determined for each block in a conversion circuit 502, and the decimal parts are rounded to become a converted input signal. A method for determining the conversion coefficient a will be explained later.

The converted input signal is encoded in the encoding device 503 in the same manner as in the first embodiment of the invention in FIG. At this time, the converted input signal is converted by the conversion circuit 502 so that it can be reversibly encoded by the encoding device 503 with a fixed word length L focus.

Therefore, since there is no need to newly obtain the divisor data OU, the encoding device 503 uses the subtracter 108 more than the encoding device shown in FIG.
, the offset generator 109 is removed, and since the converted input signal is already a signal for each block, the block divider 102 is also removed. Since the encoded data word length is constant, that is, the divisor data OU is always constant, there is no need to transmit it as an additional code.
The encoding device 503 calculates the maximum value S of the prediction error of the converted input signal.
Outputs X, the first data d1 of the block, and the encoded data. Also, conversion information from the conversion circuit 502 (inverse conversion coefficient 1
/a) is necessary for inversely converting the encoded and transmitted converted input signal into an input signal, so it is transmitted as one of the additional codes.

The conversion coefficient a is determined as follows. In the conversion circuit 502, a subtracter 510 determines the difference between the output of the predictor 509 and the input signal, that is, the prediction error. The maximum/minimum detection circuit 511 detects the maximum value SX of the prediction error for each block of the input signal.
and the minimum value SN is determined and output. From this, the dynamic range SDT of the prediction error of the input signal? becomes (SX-SN).

On the other hand, since the word length of the encoded data of the encoding device 503 is L bits, the dynamic range of the prediction error of the conversion input signal sdr (the maximum value of the prediction error of the conversion input signal is sx
, jl (5dr-sx-sn), when sn is the small value, must satisfy the following equation from the equation 0ffl.

sdr≦(2L-1) ・・・－・・・0
4) If there is no rounding error in the multiplier 514, the dynamic range s of the prediction error of the conversion input signal can be calculated by setting the conversion coefficient a to (2'-1)/(SX-SN).
dr can be set to (2'-1), and encoding can be performed efficiently. However, in reality, there are rounding errors, and 5x-R(a・SX) and sn-R(a-SN) do not necessarily hold (however, R(x) represents the value obtained by rounding the decimal part of X. ).

Therefore, the conversion coefficient a is set to (2L-1)/(SX
-SN) is set smaller. In other words, sx is R(
a-3X) or sn becomes larger than R(a-SN
), and in this case sdr is a-3
It becomes larger than DR. Since the maximum value of each error is l, the conversion information generation circuit 512 generates a conversion coefficient a that satisfies the following equation so that equation (b) is satisfied even if the influence of the rounding error becomes the worst.

(2L-3)≧R(a-SX) -R(a-SN)・
......(+51) The conversion information generation circuit 512 can be easily realized by calculating the conversion coefficient a that satisfies Equation 09 in advance and writing this into a read-only memory (ROM). Dynamic range of prediction error S
When DR satisfies the following formula, the conversion coefficient is 1.

FIG. 6 is a block diagram of a decoding apparatus according to a third embodiment of the present invention, and this decoding apparatus performs inverse transformation of the encoding apparatus shown in FIG.

In Fig. 6, 6 (11 is transformation information (inverse transformation coefficient 1/
a), 602 is the input terminal of the first data d1 in the block of the transformed input signal, 603 is the input terminal of the maximum prediction error value sx for the transformed input data, 6 (15 is the decoding of the transmitted transformed input signal) 606 is an inverse conversion circuit that inversely converts the decoded input signal into the original input signal; 212 is an inverse conversion circuit that converts the decoded input signal into the original input signal; Block decomposer 212 converts into input signals with the same data arrangement
．． 6 (17 is the output terminal of the reproduced input signal. In the decoder 6 (15, the same parts as in FIG. 3 are given the same numbers. 2 (15 is the remainder decoder, 206 is the adder, 2 (17 is a subtracter, 208 is an offset regenerator,
209 is an adder, and 210 is a switch. Inverse converter 6
06, 608 is a multiplier.

The operation of the decoding device configured as above will be explained.

The decoding device 6 (15 is basically the same configuration as the decoding device shown in Figure 3 except that the divisor data OU is not input because it does not need to be transmitted as a constant value, and the transmitted additional code The encoded data is decoded using , and becomes the original converted input signal.

The converted input signal is multiplied by the transmitted inverse transform coefficient 1/a (conversion information) in the inverse transform circuit 606 to become the original input signal. Since this input signal is a signal for each block, it is converted by the block decomposer 212 into a signal with the same data arrangement as the input signal of the encoding device, and is output from the terminal 6 (17).

As described above, in the third embodiment shown in FIGS. 5 and 6, the input signal is multiplied by the conversion coefficient a to obtain the converted input signal, and the converted input signal is converted into the first rj! The dynamic range of the prediction error of the input signal and the dynamic range of the prediction error of the converted input signal are proportional to each other, except for the influence of rounding errors. Therefore, first find the dynamic range of the prediction error of the input signal, and then set the transformation coefficients so that the dynamic range of the prediction error of the converted input signal does not exceed the dynamic range determined by the word length of the encoded data and is close to that dynamic range. Since a is set, efficient encoding is possible. In addition, the conversion coefficients are set so as not to exceed the upper limit of the dynamic range in consideration of rounding errors, so the conversion input signal is reversibly encoded and no irregular errors occur. The distortion is only due to word length limitations and rounding errors due to multiplication of transform coefficients in the conversion from the input signal to the converted input signal and its inverse conversion.

In the above example, the word length of the coded data in each block is all constant, but there is a method of making the average word length of the coded data constant for each predetermined number of blocks. Consider the case where the average word length is encoded using 4 bits. If two blocks A and B can be reversibly encoded with 6 bits and 4 bits, respectively, the average coded word length is 4 bits, so they can be encoded without deterioration. However, when the word length of encoded data is fixed to 4 bits for each block, block A suffers signal degradation of 2 bits,
The block yields 2 bits per sample. Therefore, when the transmission rate is held constant, encoding efficiency can be improved by encoding each block than when encoding each block.

In this encoding method, the following can be considered as an example of a method for determining how many bits are used to encode each block within a coding unit.

First, the dynamic range of prediction errors is calculated for all blocks within a coding unit, and the average number of bits of coded data when reversible coding is performed is determined. Conversion is performed so that the dynamic range of the prediction error of the input signal of each block becomes smaller on average by the number of bits that is the difference between this number of bits and the actual number of encoded bits.

This encoding device includes a means for determining the average number of encoded bits assuming that the encoding unit has a Batza memory and reversible encoding, and a difference between the average number of encoded bits and the actual number of encoded bits. It is sufficient to add a means for determining and storing this to the encoding device shown in FIG. 5, and this decoding device can basically be the same as the decoding device shown in FIG. 6, so a description thereof will be omitted.

Furthermore, even in the above embodiments, the dynamic range of the encoded data is not fully utilized. This is because even if the dynamic range of the prediction error expands due to rounding errors, there is a margin that does not exceed the dynamic range of the encoded data. This is because it gives rise to So the second
．． A method combining the third embodiment is conceivable. That is, in the third embodiment, a plurality of transform coefficients and a plurality of encoding devices corresponding to the transform coefficients are provided, and one that effectively utilizes the dynamic range of encoded data is selected. Second. Since the third embodiment can be easily configured, the explanation will be omitted.

Further, as a method of reducing the dynamic range of prediction errors, a method of performing nonlinear conversion as shown in FIG. 7 can be considered. In FIG. 7, the data marked with white circles are the input signals before conversion, and the data marked with black circles are the input signals after conversion.

The example in the figure is a case where the maximum prediction error value SX (=Di D+-1) occurs between the input data at time (i-1) and l, and the input data D* involved in the generation of the maximum value. −
+, Dt by +1. -I is used as the conversion input signal. As a result, the maximum prediction error S of the converted input signal
X is two levels smaller than the maximum value SX. In this way, the maximum value and minimum value of the prediction error generated by slightly decreasing or increasing some or all of the input data related to the generation of the maximum value and minimum value of the prediction error can be made slightly smaller. . That is, nonlinear processing is performed to reduce the dynamic range of prediction errors. Since the amount of change level of input data is kept small, inverse transformation at the time of decoding is not necessary. Transformation using this nonlinear processing is effective when the dynamic range of the prediction error of the input signal is equal to or slightly exceeds a power of 2, and the transformed signal is obtained by multiplying the input signal by a transformation coefficient. This is effective in suppressing an increase in prediction errors caused by rounding errors. This is because setting the dynamic range of the prediction error to a value closer to but smaller than a power of 2 means to effectively utilize the dynamic range of the encoded data, and allows the encoded data word length to be reduced by 1 bit. .

Another possible method for reducing the dynamic range of prediction errors is as follows.

Within the encoding device, the prediction error can be regarded as a high frequency component of the input signal. Therefore, a high frequency suppression filter can be considered as a method of reducing the dynamic range of prediction errors. At the time of decoding, a filter having characteristics opposite to those of the high-frequency suppression filter may be used. In this case, the conversion information may include the presence or absence of this filter and a code indicating its characteristics.

The block may have a one-dimensional configuration (for example, 166 pixels in the horizontal direction and 1 line in the vertical direction) or a two-dimensional configuration (for example, 4 pixels in the horizontal direction and 4 lines in the vertical direction), and a three-dimensional configuration that extends between fields and frames is also considered. It will be done. As a modification, a method may be considered in which after encoding is performed within a field or frame, blocks are extended in the time axis direction in units of fields or frames and then encoded. Another possible method is to make the block size variable.

In general, sample positions are continuous in both the horizontal and vertical directions! A block is composed of input data of
For signals, etc., it is also conceivable to configure a block with input data of sample positions of in-phase points of color carrier waves.

In the above embodiments, all prediction methods were pre-predictions, but the present invention is not limited to this (various prediction methods can be considered).

When the block configuration is one-dimensional, it is easiest to take out the input data in the block in the input order, and the combination of input data or decoded data used for prediction is the previous data. When it comes to two or more dimensions, there are various possibilities for the order in which input data is extracted and the combinations of data used for prediction. Figure 8 shows an example of a two-dimensional block configuration with 4 pixels in the horizontal direction and 4 lines in the vertical direction. The arrow indicates that the data indicated by the arrow is used to predict the data indicated by the arrow. Divisor data O
The smaller U is, the better the encoding efficiency is, but since it depends on the correlation of the input data, encoding is performed using multiple methods, and the output of the more efficient encoding method and the code indicating that encoding method are added to the additional code. A method of adding the information and outputting it may also be considered.

In pre-prediction, the signal with the highest coding efficiency is a monotonically increasing signal or a monotonically increasing signal. This is a signal to reduce the amount by 1 lilfi.

This is because the dynamic range of prediction errors becomes the smallest at this time. Therefore, a method can be considered in which a plurality of ways to take out input data from a block are prepared in different orders, and the one with the smallest dynamic range of prediction error is selected.

In this case, how to retrieve input data (a kind of prediction method)
A code indicating this is transmitted as one of the additional codes.

The above prediction method was extrapolation prediction, but interpolation prediction may also be used.For example, in the first embodiment, only the first data of each block is not encoded and the input data is transmitted as it is. The value obtained by linear interpolation using the leading data of each block is used as a predicted value, and the difference from the input signal is used as a prediction error. Unlike extrapolation prediction, interpolation prediction does not use sequentially decoded data to decode the next data, so error propagation does not occur.

In addition, if the high-frequency component in the input signal is very large, the input signal at a specific position in the block (for example, the first data of the block) or the average value of all data in the block can be used as the predicted signal for all input signals in the block. For example, if the input signal is 20.30.20.30, the prediction method is effective.
．． Maximum value SX of prediction error in pre-prediction in case of repetition of .

The minimum 411SN is 10. -10, and its dynamic range SDR is 20. On the other hand, if the first data 20 is a predicted value for all inputs in the same block, the maximum value Sx and minimum value SN of the prediction error are each 10.0, and the dynamic range SD
R can be as small as 10.

As described in detail, according to the present invention, it is possible to provide an encoding method and a decoding method thereof that have a simple configuration, are free from anomalous errors, and can perform efficient transmission, and have great practical value. .

[Brief explanation of the drawing]

A graph showing an example of each data in a block with a one-dimensional configuration, FIG. 3 is a block diagram showing the configuration of the decoding device in the first embodiment, and FIG. A block diagram showing the configuration of an encoding device and its decoding device, the fifth factor is a block diagram showing the configuration of a highly efficient encoding device in the third embodiment of the present invention, and FIG. A block diagram showing the configuration of the decoding device in the example, Fig. 7 is a diagram showing the order and prediction method for extracting data such as the dynamic range of prediction error of the input signal, and a diagram of operation waveforms in the method (normal state), Fig. 11 3 is a graph showing another example of an operation waveform (when an irregular error occurs) in the system. 1 (11... input terminal, 102... block division b, 103... delay device, 104...
...Register, 1 (15...Predictor, 10
6...Subtractor, 1 (17...Maximum/minimum detection circuit, 10B/old...Subtractor, 109...
Divisor data generator, 110... Remainder calculator, 1
11... Remainder encoder, 112...
Output terminal of encoded data CL, 113... Output terminal of divisor data OU, 114... Output terminal of maximum value SX of prediction error, 115... Output terminal of input data D1. Output terminal, 2 (11...encoded data C
Input terminal of i, 202... Input terminal of divisor data OU, 203 Old... Input terminal of maximum value SX of prediction error, 204... Input terminal of input data D1, 2
(15...remainder decoder, 206...
Adder, 2 (17... Subtractor, 208...
・Offset regenerator, 209 ・Old ・Offset adder, 210 ・Switch, 211 ・Old ・Predictor, 212 ・Block decomposer, 213 ・
...Output terminal for decoded data. Ward tjlm-! Bep Figure Ft :ou-NIS D+ <OLJ・(N+10 data) No. 1 t! 11n-magazine rho 7th mouth t-s mu-2t-t A i middle I A middle 2
Figure (Cν (eJ C9) (d) (f) (h) Pot diagram Figure 10

Claims (27)

[Claims] (1) A block consisting of a plurality of input data is used as a unit of encoding, and the maximum value SX and minimum value SN of the prediction error, which is the difference between the input data and its predicted value in the block, are determined, and (SX-SN ) is the larger predetermined value as the divisor data O
When U, the remainder E obtained by dividing each input data of the block by the divisor data OU is encoded, and the encoded E is combined with an additional code containing prediction error information regarding the SX, SN, and OU. An encoding method characterized by transmitting both. (2) The encoding method according to claim (1), wherein the predicted value is input data in the same block as the predicted data. (3) The encoding method according to claim (1), wherein the predicted value is input data at a specific position within the same block as the predicted data. (4) Claim (1) characterized in that the predicted value is an average value of input data in the same block as the predicted data.
) Encoding method described. (5) The encoding method according to claim (1), which uses interpolation prediction. (6) The encoding method according to claim (1), wherein prediction information indicating a prediction method with the highest encoding efficiency is used as one of the additional codes. (7) Divisor data O transmitted as one of prediction error information
2. The encoding method according to claim 1, wherein U satisfies the following formula: OU=2^M>SX-SN≧2^M^-^1, where M is an integer. (8) The encoding method according to claim (1), wherein input data at a specific position within a block is one of the additional codes. (9) The encoding method according to claim (1), wherein the average value of input data within a block is used as one of the additional codes. (10) The encoding method according to claim (1), wherein the input signal is a signal obtained by converting the original input signal so that the dynamic range of the prediction error of the original input signal is equal to or less than a predetermined amount for each predetermined number of blocks. (11) Claim (1) characterized in that conversion information indicating a method of converting the original input signal is transmitted as one of the additional codes.
0) Encoding method described. (12) The encoding method according to claim (10), wherein the method for converting the original input signal comprises multiplying the original input signal by a predetermined coefficient. (13) The encoding method according to claim (10), wherein the method of converting the original input signal is to increase or decrease a part of the original input signal by a minute amount. (14) Claim (10) characterized in that the method for converting the original input signal is to perform high frequency suppression on the original input signal.
) Encoding method described. (15) Using a block composed of multiple input data as a unit of encoding, find the maximum value SX and minimum value SN of the prediction error, which is the difference between the input data in the block and its predicted value, and (SX-SN ) When a larger predetermined value is set as the divisor data OU, the remainder E obtained by dividing each input data of the block by the divisor data OU is encoded, and includes prediction error information regarding the SX, SN, and OU. A signal encoded by a coding method that transmits both an additional code and the encoded E is input, the divisor data OU is obtained from the transmitted additional code, and an integer N is added to the transmitted and decoded remainder E. The integer N is calculated by adding the divisor data OU times the divisor data OU to obtain the decoded data, and the integer N A decoding method characterized by determining. (16) The decoding method according to claim 15, wherein the predicted value is input data in the same block as the predicted data. (17) The decoding method according to claim (15), wherein the predicted value is input data at a specific position within the same block as the predicted data. (18) Claim characterized in that the predicted value is an average value of input data in the same block as the predicted data (
15) The decoding method described. (19) The decoding method according to claim (15), which uses interpolation prediction. (20) The decoding method according to claim (15), wherein prediction information indicating a prediction method with the highest encoding efficiency is used as one of the additional codes. (21) Claim characterized in that the divisor data OU transmitted as one of the prediction error information satisfies the following formula OU=2^M>SX-SN≧2^M^-^1, where M is an integer ( 15) The decoding method described. (22) The decoding method according to claim (15), wherein input data at a specific position within the block is one of the additional codes. (23) The decoding method according to claim (15), wherein one of the additional codes is an average value of input data within a block. (24) The decoding method according to claim (15), wherein the original input signal is reproduced by performing the reverse conversion on the decoded data based on the conversion information in the additional code. (25) A claim characterized in that the decoding of the encoded data after an error occurs in the encoded data is performed by predicting from a direction different from that at the time of encoding using the decoded data in an adjacent block as the starting point of prediction. The decoding method described in item (15). (26) Claim (15) characterized in that the additional code in which an error has occurred is decoded using the additional code of an adjacent block.
) Decryption method described. (27) If there is a region where the prediction error obtained by performing prediction across block boundaries using decoded data is outside the range of the prediction error determined from the additional code, it is determined that a transmission error has occurred, and 16. The decoding method according to claim 15, wherein the decoded area is predicted and decoded from a direction different from that during encoding using decoded data as a starting point for prediction.