JPH01303871A - Method and apparatus for highly efficient encoding - Google Patents

Abstract

PURPOSE:To obtain a decoding level corresponding well to an original analog level by obtaining a dynamic range with the width of the original analog level taken into account and applying encoding/decoding in response to the dynamic range. CONSTITUTION:A code signal Qi and a decoding level Li are obtained by the processing expressed in equation I. That is, the dynamic range D corresponding to the analog quantity is obtained from a maximum value and a minimum value in a block as the original analog quantity, and the dynamic range D is split uniformly into 2<n> to obtain a quantized step width (d). Then the analog minimum value is eliminated and the remaining value is divided by the quantizing width (d) to obtain the code signal Qi in an area to which the digital value Li belongs, that is, by the re-quantization depending on integral number processing due to truncation. Moreover, the decoded value Li is obtained by rounding the median of each level range. Thus, the decoded level is made well correspondent with the original analog level.

Description

[Detailed Description of the Invention] [Field of Industrial Application] This invention is a high-efficiency code that is applied to compress the data amount of a digital image signal, transmit it, and restore the original image data on the receiving side. The present invention relates to a method and an apparatus for oxidation.

[Summary of the invention]

In this invention, a dynamic range defined by the maximum and minimum values of a plurality of pixel data included in a block of a digital image signal is detected, the dynamic range is divided into a plurality of level ranges, and the minimum value is removed. Check to which level range the pixel data or the level difference of the pixel data with respect to the maximum value belongs, and correspond to the level range to which it belongs,
In a high-efficiency encoding device that generates a code signal with a smaller number of bits than the original number of bits, and obtains a restoration level corresponding to the original analog level from the dynamic range information and the code signal on the receiving side, By determining the dynamic range in consideration of the width of the original analog level and performing encoding and decoding in accordance with this dynamic range, a restored level that closely corresponds to the original analog level can be obtained.

[Conventional technology]

The applicant of this application obtains a dynamic range defined by the maximum value and minimum value of a plurality of pixel data contained in a two-dimensional block, as described in Japanese Patent Application No. 59-266407, and calculates this dynamic range. We have proposed a high-efficiency encoding device that performs range-adaptive encoding. Furthermore, as described in Japanese Patent Application No. 60-232789, a high-efficiency encoding device performs encoding adapted to a dynamic range with respect to a three-dimensional block formed from pixels in areas included in each of a plurality of frames. is proposed. Furthermore, in the specification of Japanese Patent Application No. 60-268817,
A high-efficiency encoding device for variable-length codes is shown in which the number of bits changes depending on the dynamic range so that the maximum distortion that occurs when quantization is constant.

A previously proposed encoding device adapted to a dynamic range will be described with reference to the drawings.

FIG. 1 shows the configuration of the transmitting side, and when applied to a digital VTR, the configuration of the recording side.

In FIG. 1, a digital video signal in which one sample is quantized to 8 bits, for example, is supplied to an input terminal indicated by 1. This digital video signal is sent to the blocking circuit 2.
The blocking circuit 2 converts the input digital video signal into continuous signals in units of two-dimensional blocks, which are units of encoding. In the blocking circuit 2, the second
As shown in Figure B, one frame of the screen is from Bll to Bmn.
subdivided into a large number of blocks. 1st block is 2nd block
As shown in Figure A, the size is, for example, (4 lines x 4 pixels).

The output signal of the blocking circuit 2 is supplied to a maximum value detection circuit 3, a minimum value detection circuit 4, and a delay circuit 5. The maximum value detection circuit 3 detects the maximum value Lwax for each block, and the minimum value detection circuit 4 detects the minimum value Latin for each block. The maximum value L max from the maximum value detection circuit 3 and the minimum value La1n from the minimum value detection circuit 4 are supplied to the subtraction circuit 6, and a dynamic range D represented by (Lmax - Lmin=D) is detected.

Furthermore, the pixel data Li and the minimum value L sin passed through the delay circuit 5 are supplied to the subtraction circuit 7, and the subtraction circuit 7 obtains data after the minimum value has been removed, which is represented by (Li −L+sin ). The data after the minimum value has been removed from the subtraction circuit 7 is supplied to an ADRC (encoding adapted to dynamic range) encoder 8. The encoder 8 is supplied with a dynamic range D, and the encoder 8 generates a code signal Qi of n bits, which is smaller than the original number of bits.

The dynamic range D, the minimum value Lnin, and the code signal Qi are supplied to the framing circuit 9, subjected to error correction encoding processing, and converted into transmission data. Transmission data is obtained at the output terminal 1° of the framing circuit 9, and this transmission data is transmitted.

The receiving side (in the case of a digital VTR, the reproducing side) has the configuration shown in FIG. In FIG. 3, received data is supplied to an input terminal indicated by 11, and a frame decomposition circuit 12 decodes an error correction code and separates a dynamic range D, a minimum (!La1n, and a code signal Qi). Dynamic range D and code signal Qi are supplied to an ADRC decoder 13, and the output signal of the decoder 13 is supplied to an adder circuit 14.

In the adder circuit 14, the output signal of the decoder 13 and the minimum value L
Ilin is added to obtain the restoration level Li. This restoration level li is supplied to the block decomposition circuit 15,
Block order is converted to traversal order. Output terminal 16
, the reconstructed data is returned to the scanning order.

In the ADRC encoder 8 and decoder 13 described above,
Encoding and decoding shown in FIG. 4A or FIG. 4B are performed.

The quantization shown in Figure 4A divides the dynamic range D equally into (2n) (4 level ranges in this example), and assigns a code depending on which level range the data belongs to after the minimum value is removed. This quantization method is
This is called non-edge matching. In non-edge matching, the code signal Qi is (0-00) (1=
01) (2-10) (3=11) is obtained, and the median value of each level range is taken as the restoration level (0, 1, 2, or 3).

The quantization shown in Figure 4B divides the dynamic range D equally into (2n-1) (in this example, 3 level ranges), then shifts the divided level ranges to Encoding is performed depending on which level range the data after minimum value removal belongs to among the four level ranges obtained.
Such a quantization method is called edge matching. In edge matching, the same code signal Qi as in non-edge matching is obtained, and the restoration level ( 0.1゜2
or 3), edge matching has a larger quantization step width than non-edge matching, but
The minimum value L win and maximum value Lwax within the block can be restored with zero error. This means that the dynamic range of the block is preserved, and when encoding and decoding operations are repeated, such as in the dubbing operation of a VTR, deterioration in image quality is prevented.

In conventional ADRC, both non-edge matching and edge matching are encoded (requantized) according to the following formula.
And decryption was performed.

D = Lmax - La+in In order to obtain the above restoration levels f and l, Qi is set to 0.5.
is added in order to set the median value of the divided level range as the restoration level, and also adds 0.5 to the calculated value.
Adding 0.5 is equivalent to rounding off without adding 0.5.

[Problem to be solved by the invention]

As an example, as shown in FIG.
, Lmax-84) is requantized by 2 bits by non-edge matching. Considering the 80 level of the minimum value L + win, this level is (7) at the original analog level, as shown by the circle.
9,50... to 80.49...) are representative. That is, the original information level included in this block is (84, 49-79, 50-5, 0). Therefore, if we consider the original analog level faithfully, the dynamic range D is (D= However, in conventional encoding, since +1 is not added, a restoration level that corresponds well to the original analog level cannot be obtained.

On the other hand, when encoding with edge matching, the maximum value 1 wax and the minimum value l 1li as analog values
Let n be the maximum value Ls+ax and the minimum value Lwin in the block.
Therefore, the dynamic range D is (D=lt*
ax −1aax −Ln+ax −1,+in ), and the analog dynamic range matches that of the digital one. Therefore, it is not necessary to perform the (+1) processing as described above.

Therefore, an object of the present invention is to provide a highly efficient encoding method and apparatus that can obtain a restoration level that corresponds well to the original analog level for both non-edge matching and edge matching. be.

[Means to solve the problem]

In this invention, a dynamic range defined by the maximum and minimum values of a plurality of pixel data included in a block of a digital image signal is detected, and the dynamic range is divided into 2n level ranges obtained by equally dividing the dynamic range. Based on the level relationship with the pixel data from which the minimum value has been removed, the pixel data is encoded with a smaller number of bits n than the original number of bits, that is, encoded by non-edge matching, and the information obtained by dynamic range information and encoding is In a high-efficiency encoding method in which a restoration level is obtained from a code signal, the code signal Qi and restoration level Li are
It is done so as to obtain.

D = Lmax -1, +in + 1 However, Lma
x Maximum value of Ni block 1. min Minimum value of two blocks Furthermore, in the present invention, the dynamic range defined by the maximum and minimum values of a plurality of pixel data included in a block of a digital image signal is detected, and the dynamic range is calculated as (2n-1)( 2''' obtained by shifting the level range evenly divided into n: number of bits of code signal) by %
Based on the relationship between the level range and the pixel data from which the minimum value has been removed, the pixel data is encoded with a smaller number of bits than the original number of bits, that is, encoded by edge matching, and the dynamic range information and the information obtained by encoding are encoded using edge matching. In a high-efficiency encoding method in which a restoration level is obtained from a code signal, the code signal Qi and restoration level lj are obtained by the following processing.
It is done so as to obtain.

D = Ltsax -La1n Qi=+a However, L, tsax Ni block maximum value La1n
Furthermore, in this invention, the minimum value L is determined from the value of each pixel data.
Unlike encoding data with 5hin removed,
The same applies to the case where the difference between the maximum value LIIaχ and the value of each pixel data is encoded by non-edge matching or edge matching.

〔Example〕

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

This description is given in the following order.

a. One embodiment (non-edge matching) b. Another embodiment (edge matching) C0 modification a. One embodiment (non-edge matching) -80, Lma
x -84) The maximum value 1 taax and the minimum value 11 win in the block as 0-element analog quantities
is Lmax=Lo+ax+0. 5! ! , nin −Lain +0. 5 Therefore, the dynamic range D corresponding to the analog quantity is D = Lmax - I! , min = Lmax −
Lmin +1.0. In n-bit quantization, the quantization step width d is determined by equally dividing this dynamic range D into 2n parts.

In the case of d-D/2n (n-2), as shown in Figure 6, (5,0
/4-1.25). To find out which level range the digital value Li of each pixel belongs to, consider this digital (direct Lt) to be the median value of the range of analog quantities of ±0.5.Then, the minimum analog value 1m1n (
By removing the offset) and dividing by the quantization width d, the area to which the digital value Li belongs can be found by rounding down to an integer. That is, the code signal Qi (0
~2n-1) is obtained in the case of @(n-2),
Code signals Qi (0 to 3) are obtained.

As mentioned above, since areas are divided at the analog level, ±LSB (least significant bit)/2 of the digital image
It will be treated more broadly. Therefore, in the example shown in FIG. 5, as shown in FIG.
corresponds to. The quantization code Qi is determined by the following equation.

In the above formula, [ ] means taking the integer part by rounding down. Lmax and Ln+in are each 2n
quantized to -1 and 0. However, an intermediate digital level may coincide with a boundary of a level range, and a digital level that coincides with this boundary is included in the upper level range.

In order to realize the above-mentioned encoding, for example, the configuration shown in FIG. 8 is used. (Lwax −La
The output signal of tin) is supplied to the +1 adder circuit 21,
A dynamic range D is obtained from the +1 adder circuit 21. In the case of quantization with 2 bits, the dynamic range D is supplied to an

Further, an output signal of (Li −Lmtn ) from the subtraction circuit 7 is supplied to the division circuit 23 via the +0.5 addition circuit 24 . Therefore, the division circuit 23 performs the operation in [ ] in the above equation. The output signal of the division circuit 23 is supplied to the integer conversion circuit 25, and the code signal Qi is obtained by rounding down the signal in the integer conversion circuit 25.

The configuration of the encoder 8 shown in FIG. 8 described above is an example, and other configurations such as a combination of a level comparison circuit and a priority encoder, a configuration using a ROM, software processing, etc. are possible. Furthermore, when transmitting dynamic range information, it is sufficient to transmit any two data of the dynamic range D, the maximum value LavaX, and the minimum value La1n.

Next, the restored value (digital value to be decoded) 1i is
As shown in the formula below, the median value of each level range is rounded off.

i, i-(dXQi +'Ad+i, win) Rounding = (dXQi +'Ad+fa+in +0.5
) Truncation = (d X (Qi +0.5) +Lm
in ) Truncation An example of the configuration for performing the above decoding is shown in Part 9.
The received code signal Qi shown in the figure is supplied to a multiplication circuit 32 via a +0.5 addition circuit 31. In addition, the dynamic range D is transmitted to the multiplication circuit 32 via the
The multiplier circuit 32 performs the calculations in brackets [ ] in the above equation. The output signal of this multiplier circuit 32 is
4 and undergoes truncation processing. Integer conversion circuit 3
The output signal of 4 is supplied to the adder circuit 14, and the adder circuit 14
By adding the minimum value LLIin, the restoration level li is obtained.

As the decoder 13, in addition to the configuration shown in FIG.
It is possible to configure using , software processing, etc.

b. Other Examples (Edge Matching) Next, another example in which the present invention is applied to edge matching will be described. Similar to the above embodiment,
Consider discrete digital image levels as appropriate for analog continuous quantities.

The purpose of edge matching is to restore the maximum value LmaX and minimum value Lmin within a block with an error O. For this purpose, the maximum value 1 t as an analog level
aax and minimum value j! win sets the digital value itself.

lrmax=Lmax ffimin=Lmin Therefore, as shown in FIG. 10, the analog dynamic range matches that of the digital one.

D-Ila+ax-Ilwin-1,+ax-L
ain If this dynamic range D is equally divided to include the levels at both ends, the quantization step width d will be 2n-12·'-1.

As shown in Fig. 11, the code signal Qi is obtained by shifting the difference from the analog minimum value fmin by the quantization step width d (0 to 3 on the left in Fig. 11) by 0.5. (O to 3 on the right side in FIG. 11), which is a uniform area. Therefore, encoding in the case of edge matching is performed by the following equation.

The restoration level L, i is obtained by rounding off an analog value that is Qi quantization step widths away from the analog minimum value fmin. That is, i, i −(Qi Xd
+1m1n +0.5) The restoration level li is determined by rounding down.

Encoders and decoders according to other embodiments of the present invention can be realized using arithmetic circuits, ROM, software processing, etc., similarly to the above-described embodiment.

C6 Modified Example The above description is an example in which the present invention is applicable to ADRC which quantizes a value obtained by removing the minimum value La1n of a block from data Li of each pixel.

However, the present invention can be applied to ADRC in which the difference between the maximum value L wax of a block and the data Li of each pixel is detected and this difference is encoded by an encoder.

That is, encoding and decoding in the case of non-edge matching are performed using the following formula.

D = Lmax - Lmin + 1 In the case of edge matching, encoding and decoding are performed using the following formula.

D = Lmax - Lmin In the above embodiments, encoding and decoding are performed using either non-edge matching or edge matching. However, a configuration may be adopted in which either of the two methods is selectively activated depending on the purpose and the like. Furthermore, the present invention is applicable not only to ADRC having a two-dimensional block structure but also to a case where a three-dimensional block is constructed from a plurality of regions each included in a plurality of temporally continuous frames.

〔Effect of the invention〕

According to this invention, when dividing the dynamic range D defined by the maximum and minimum values detected in a block of a digital image signal into a plurality of level ranges, the width of the original analog level is taken into account. Therefore, the restored level corresponds well to the original analog level.

[Brief explanation of the drawing]

Figure 1 is a block diagram of the configuration of the transmitting side of a high-efficiency encoding device to which this invention can be applied, Figure 2 is a route diagram used to explain the blocks, and Figure 3 is a high-efficiency encoding device to which this invention can be applied. Fig. 4 is a route diagram used to explain quantization, Fig. 5 is a route diagram used to specifically explain quantization, and Figs. 6 and 7 are block diagrams of the configuration of the receiving side of this invention. The route map used to explain one embodiment, FIG. 8 is a block diagram of an example of an encoder to which this invention is applied, and the ninth part is a block diagram of a decoder to which this invention is applied, and FIGS.
FIG. 1 is a block diagram used to explain another embodiment of the invention. Explanation of main symbols in the drawings 2 Niblock conversion circuit, 3: Maximum value detection circuit, 4: Minimum value detection circuit, 8: Encoder, 13: Decoder, 15 Niblock decomposition circuit, 21: +1 addition circuit, 23: Division calculation circuit, 24.31: +0.5 addition circuit, 32: multiplication circuit. Agent Patent Attorney Masato SugiuraFigure 2Figure 3Figure 4Figure 8Figure 1'3-7Figure 7. Fig. 10 Fig. 6 84.5 -- 93.25 -- 32.0 -- 80.75 -- 79.5 -- jil1 diagram

Claims (8)

[Claims] (1) 2^n levels obtained by detecting a dynamic range defined by the maximum and minimum values of multiple pixel data included in a block of digital image signals and equally dividing the dynamic range. Based on the level relationship between the range and the pixel data from which the minimum value has been removed, the pixel data is encoded with a number n of bits smaller than the original number of bits, and from the information on the dynamic range and the code signal obtained by the encoding. A high-efficiency encoding method for obtaining a restoration level, characterized in that the code signal Qi and the restoration level ■i are obtained by the following processing. D=Lmax-Lmin+1 Qi=[(Li-Lmin+0.5)×2^n/D] Round down ■i=[D×(Qi+0.5)/2^n+Lmin] Round down However, Lmax: Maximum value of the block Lmin: Minimum value of the block (2) 2^n levels obtained by detecting a dynamic range defined by the maximum and minimum values of multiple pixel data included in a block of a digital image signal and equally dividing the dynamic range. Based on the level relationship between the range and the pixel data from which the minimum value has been removed, the pixel data is encoded with a number n of bits smaller than the original number of bits, and from the information on the dynamic range and the code signal obtained by the encoding. A high-efficiency encoding device for obtaining a restoration level, characterized in that the code signal Qi and the restoration level i are obtained by the following processing. D=Lmax-Lmin+1 Qi=[(Li-Lmin+0.5)×2^n/D] Round down Li=[D×(Qi+0.5)/2^n+Lmin] Round down, where Lmax: Maximum value of the block Lmin :Minimum value of block (3) 2^n levels obtained by detecting a dynamic range defined by the maximum and minimum values of multiple pixel data included in a block of a digital image signal and equally dividing the dynamic range. Based on the relationship between the range, the maximum value, and the level difference of the pixel data, the pixel data is encoded with a number n of bits smaller than the original number of bits, and restored from the dynamic range information and the code signal obtained by the encoding. A high-efficiency encoding method characterized in that the code signal Qi and the restoration level i are obtained by the following processing. D=Lmax-Lmin+1 Qi=[(Lmax-Li+0.5)×2^n/D] Round down Li=[Lmax-D×(Qi+0.5)/2^n] Round down, where Lmax: Maximum of the block Value Lmin: Minimum value of the block (4) 2^n levels obtained by detecting a dynamic range defined by the maximum and minimum values of multiple pixel data included in a block of digital image signals and equally dividing the dynamic range. Based on the relationship between the range, the maximum value, and the level difference of the pixel data, the pixel data is encoded with a number n of bits smaller than the original number of bits, and restored from the dynamic range information and the code signal obtained by the encoding. A high-efficiency encoding device characterized in that the code signal Qi and the restoration level i are obtained by the following processing. D=Lmax-Lmin+1 Qi=[(Lmax-Li+0.5)×2^n/D] truncation ■i=[Lmax-D×(Qi+0.5)/2^n] truncation However, Lmax: of the block Maximum value Lmin: Minimum value of block (5) Detect the dynamic range defined by the maximum and minimum values of a plurality of pixel data included in a block of digital image signals, and convert the dynamic range to (2^
From the relationship between the 2^n level ranges obtained by 1/2 shifting the level range equally divided into n-1) (n: the number of bits of the code signal) and the pixel data from which the above minimum value has been removed. In a high-efficiency encoding method in which the above pixel data is encoded with a smaller number of bits than the original number of bits, and a restoration level is obtained from the above dynamic range information and the code signal obtained by the above encoding, the following processing is performed. A highly efficient encoding method characterized in that the code signal Qi and the restoration level i are obtained by: D=Lmax-Lmin Qi= [(Li-Lmin)(2^n-1)+0.5/D] Round down ■i=[D×Qi/(2^n-1)+Lmin+0.5
] Round down However, Lmax: Maximum value of block Lmin: Minimum value of block (6) Detect the dynamic range defined by the maximum and minimum values of a plurality of pixel data included in a block of digital image signals, and convert the dynamic range to (2^
From the relationship between the 2^n level ranges obtained by 1/2 shifting the level range equally divided into n-1) (n: the number of bits of the code signal) and the pixel data from which the above minimum value has been removed. In a high-efficiency encoding device that encodes the pixel data with a smaller number of bits than the original number of bits and obtains a restoration level from the dynamic range information and the code signal obtained by the encoding, the following processing is performed. A high-efficiency encoding device characterized in that the code signal Qi and the restoration level i are obtained by: D=Lmax-Lmin Qi= [(Li-Lmin) (2^n-1)/D+0.5] Round down ■i=[D×Qi/(2^n-1)+Lmin+0.5
] Round down However, Lmax: Maximum value of block Lmin: Minimum value of block (7) Detect the dynamic range defined by the maximum and minimum values of a plurality of pixel data included in a block of digital image signals, and convert the dynamic range to (2^
Based on the relationship between the 2^n level range obtained by shifting the level range equally divided into n-1) (n: the number of bits of the code signal), the above maximum value, and the level difference between the above pixel data, the above pixel In a high-efficiency encoding method in which data is encoded with a smaller number of bits than the original number of bits, and a restoration level is obtained from the dynamic range information and the code signal obtained by the encoding, the following processing is performed. A highly efficient encoding method characterized in that the code signal Qi and the restoration level i are obtained. D=Lmax-min Qi= [(Lmax-Li)(2^n-1)/D+0.5] Round down ■i=[Lmax-D×Qi/(2^n-1)+0.5
] However, Lmax: Maximum value of the block Lmin: Minimum value of the block (8) Detect the dynamic range defined by the maximum and minimum values of a plurality of pixel data included in a block of digital image signals, and convert the dynamic range to (2^
From the relationship between the 2^n level range obtained by 1/2 shifting the level range equally divided into n-1) (n: the number of bits of the code signal), the above maximum value, and the level difference between the above pixel data. In a high-efficiency encoding device that encodes the pixel data with a smaller number of bits than the original number of bits and obtains a restoration level from the dynamic range information and the code signal obtained by the encoding, the following processing is performed. A high-efficiency encoding device characterized in that the code signal Qi and the restoration level i are obtained by: D=Lmax-Lmin Qi= [(Lmax-Li)(2^n-1)/D+0.5] Round down ■i=[Lmax-D×Qi/(2^n-1)+0.5
] Round down. However, Lmax: Maximum value of block Lmin: Minimum value of block