JPH07288809A - Coder for digital picture signal - Google Patents

Abstract

PURPOSE:To attain a high compression rate by controlling the fineness of quantization by both luminance and activity so as to suppress deterioration in the image quality when coefficient data generated by cosine transformation is quantized. CONSTITUTION:AC coefficient data and DC coefficient DCT are generated from a DCT transmission circuit 4. The AC coefficient data are quantized by a quantization circuit 7 and subjected to variable length coding by a VLC encoder 11. A quantization number controller 10 detects the degree of luminance from the DC coefficient data of a Y signal and classifies the data into macro blocks based on a maximum value of the absolute value of the AC coefficient data. A data quantity estimate device 9 selects a quantization number in the quantization table for each of 5 macro blocks and it is fed to the quantization number controller 10 and the quantization number is adjusted by class information. A class of most rough quantization is allocated to a macro block whose luminance is high to make the deterioration in the image quality unremarkable.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital image signal coding apparatus using a cosine transform.

[0002]

2. Description of the Related Art A digital VTR for recording a digital video signal on a magnetic tape by a rotary head is known. Since the amount of information in a digital video signal is large,
High-efficiency coding for compressing the amount of transmitted data is often adopted. Among various high efficiency coding, DC
Practical application of T (Discrete Cosine Transform) is progressing.

In the DCT, one frame image is converted into, for example, (8
X8) is converted into a block structure, and this DCT block is subjected to cosine transform processing which is a kind of orthogonal transform. As a result, (8 × 8) coefficient data is generated. Such coefficient data is quantized, subjected to variable length coding processing such as run-length coding and Huffman coding, and then transmitted. At the time of transmission, in order to facilitate data processing on the reproduction side, a code signal, which is an encoded output, is inserted into the data area of a sync block of a certain length, and a sync signal and an ID signal are added to the code signal. The frames that form the sync blocks are framed.

It is considered that, when performing quantization, in order to improve the compression efficiency while suppressing the deterioration of the image quality of the restored image, the quantization process is performed according to the feature (activity) of the image. The applicant of the present application proposes to classify the DCT blocks into classes according to the activity and quantize the coefficient data in a quantization step according to the class. This is because an image with high activity has no noticeable visual distortion even if the quantization step is roughened. In the method proposed earlier, as a class classification according to activity,
The maximum absolute value of the AC (alternating current) coefficient was used.

[0005]

In the classification method as described above, there are still cases where even if coarse quantization is performed, fine quantization can be performed although deterioration is not noticeable. As a result, there is room for further improvement in compression efficiency.

Therefore, an object of the present invention is to provide a coding apparatus for a digital image signal, which can improve the previously proposed adaptive quantization of DCT coefficients and can further improve the efficiency.

[0007]

According to the present invention, there is provided a digital image signal coding apparatus, wherein a digital image signal is coded by cosine transform, and AC coefficient data generated by the coding is variable length coded. The quantization circuit for quantizing the coefficient data and the quantization table prepared in advance are used to refer to the quantization table of the quantization circuit in order to keep the data amount of the variable length coded output for a predetermined period below the target value. A data amount estimation circuit for selecting the quantization step and for receiving the DC coefficient data and the AC coefficient data generated by the encoding, and determining the class indicating the fineness of the quantization of the partial image to be quantized. And a circuit for adjusting the quantization step for a predetermined period determined by the data amount estimation circuit according to the class of the partial image. An encoding apparatus in a digital image signal.

[0008]

When the width of the quantization step is controlled according to the characteristics of the partial image, the brightness of the partial image is detected from the DC component of the coefficient data obtained by the cosine transform. When the brightness is high, the deterioration due to the coarse quantization step can be made visually inconspicuous. As a result of the margin generated by roughening the quantization step, other data can be quantized more finely.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows the configuration of a video data processing circuit provided on the recording side of a digital VTR. Figure 1
In, the digitized video data is supplied to the input terminal indicated by 1. This video data is supplied to the blocking circuit 2. In the blocking circuit 2, the video data in the interlace scanning order is, for example, (8 × 8).
Is converted into data of the DCT block structure. That is, two (4 × 8) blocks at the same spatial position in the first and second fields that are temporally consecutive are combined to form an (8 × 8) block. In the (8 × 8) block, pixel data on odd-numbered lines is included in the first field, and pixel data on even-numbered lines is included in the second field.

The output of the blocking circuit 2 is supplied to the shuffling circuit 3. In the shuffling circuit 3, to prevent errors from concentrating due to dropouts, scratches on the tape, head clogs, etc., and conspicuous deterioration of image quality,
Within one frame, with multiple macroblocks as a unit,
The process of making the spatial position different from the original one, that is,
Shuffling is done. In this example, the shuffling unit is equal to the buffering unit and is set to 5 macroblocks.

The output of the shuffling circuit 3 is supplied to a DCT (cosine transform) circuit 4 and a motion detection circuit 5.
From the DCT circuit 4, (8 × 8) coefficient data (that is, coefficient data of DC component AC and AC component AC) is generated. This DCT circuit 4
Switching is performed so that the intra-field DCT is performed on the (4 × 8) block included in the (8 × 8) block.

A macroblock is a collection of a plurality of (8 × 8) coefficient data per DCT block. For example, in the case of (Y: CB: CR = 4: 1: 1) video data of the component system of the 525/60 system, as shown in FIG. 2A, four Y blocks at the same position in one frame And 1 CB block and 1 C
A total of 6 blocks including R blocks constitute one macroblock. When the sampling frequency is 4 fsc (fsc: color subcarrier frequency), one frame image is (9
10 samples × 525 lines), and the valid data therein is (720 samples × 480 lines). In the case of the component system described above, the total number of blocks in one frame is (720 × 6/4) × 480 / (8 × 8)
= 8100. Therefore, 8100 ÷ 6 =
1350 is the number of macroblocks in one frame.

In the case of (Y: CB: CR = 4: 2: 0) video data of the component system of the 625/50 system, as shown in FIG. 2B, four video data at the same position in one frame are used. A total of 6 blocks including the Y block, one CB block and one CR block constitute one macro block.

The DC (direct current component) coefficient data of the (8 × 8) coefficient data generated in the DCT circuit 4 is transmitted to the circuit in the subsequent stage without being compressed, and the remaining 63 pieces of AC coefficient data are buffered by the buffer 6. Is supplied to the quantization circuit 7 via. AC
As shown in FIG. 3, the coefficient data is transmitted in the order of zigzag scanning from an alternating current component having a lower order to one having a higher order. Further, this AC coefficient data is also supplied to the classification circuit 8 and the data amount estimator 9. The buffer 6 delays the coefficient data for a time required for the estimator 9 to determine an appropriate quantization number QNo, and
It is provided to output the coefficient data of each of the still block and the moving block in a predetermined order. The quantization number QNo from the estimator 9 is supplied to the quantization circuit 7 and transmitted to the subsequent stage.

The above-mentioned generation of the coefficient data from the DCT circuit 4 is the case of the DCT conversion within the frame,
When the motion detection circuit 5 detects that there is motion,
The DCT process in the field is selected. That is,
The intra-field DCT is to perform DCT for every two (4 × 8) blocks at the same position in the first and second fields that are temporally continuous. If the motion detection circuit 5 detects that there is a motion between fields in the block, the intra-frame DCT is changed to the intra-field DCT in response to the detection. The motion detection circuit 5 displays (8
For each block, the still / movement determination is performed based on the coefficient data in the vertical direction when the image data of the block of (8) is Hadamard transformed. Alternatively, the motion detection may be performed based on the absolute value of the field difference.

In the case of the intra-field DCT, (4 × 8) coefficient data for the first field and (4 × 8) coefficient data for the second field are generated,
These are located above and below (8 x
It is treated as an array of 8). The DC data DC1 is included in the coefficient data of the first field. Similarly, the second field also contains the DC component DC2. If the coefficient data of each of these fields is treated separately, D
The subsequent processing is unavoidable separately for the CT and the in-field DCT. As a result, problems such as an increase in the scale of hardware occur. Therefore, the DC component D of the second field
Instead of C2, the differential DC component ΔDC2 (= DC1-DC
2) is transmitted.

The detection signal (motion flag) M from the motion detection circuit 5 is supplied to the data amount estimator 9 and is inserted into the recording data at the subsequent stage. In the data amount estimator 9, the motion flag M is used to switch the output order of coefficient data and the area division method depending on the stillness / motion.

In the quantizing circuit 7, the AC component in the coefficient data is quantized. That is, A with an appropriate quantization step
The C coefficient data is divided and the quotient is converted into an integer. This quantization step is determined by the quantization number QNo from the QNo controller 10. In the case of a digital VTR, since processing such as editing is performed in units of one field or one frame, it is necessary that the amount of generated data in one field or one frame be equal to or less than the target value.
The amount of data generated by DCT and variable-length coding varies depending on the pattern to be coded. Therefore, a buffering process for reducing the amount of generated data in a buffering unit shorter than one field or one frame period to a target value or less. Is done. The reason for shortening the buffering unit is to simplify the buffering circuit, such as reducing the memory capacity for buffering. In this example, 5
A macro block (= 30 DCT block) is a buffering unit.

As will be described later, the classifying circuit 8 checks the fineness of the picture in units of macroblocks, classifies the activity of the macroblocks into four levels, and sets a 2-bit class indicating the class. The activity code AT is generated. The detection result is supplied to the QNo controller 10, and the activity code AT is inserted in the recording data in the subsequent stage.

The output of the quantizing circuit 7 is the variable length coding circuit 1.
1, run-length coding, Huffman coding, etc. are performed. For example, a run length, which is the number of consecutive zeros of coefficient data, and the value of coefficient data are given to a Huffman table stored in ROM, and a variable length code (encoded output) is given.
Two-dimensional Huffman coding is used to generate The code signal from the variable length coding circuit 11 is supplied to the subsequent stage.

In connection with the estimator 9, the same Huffman table 12 referred to by the variable length coding circuit 11 is provided. The Huffman table 12 generates bit number data of an output code when variable length coding is performed.
The estimator 9 determines the optimum set of quantization steps, and the determination output is supplied to the QNo controller 10. Q
The No controller 10 controls the quantization circuit 7 to quantize the coefficient data in the set of the quantization steps.
At the same time, the quantization number QNo for identifying the set of quantization steps is transmitted to the subsequent stage.

Although not shown, the data (DC coefficient DCT, variable length coded output, quantization number QN)
o, the motion flag M, and the activity code AT) are subjected to error correction coding processing and conversion of recording data into a frame structure in a framing circuit in the subsequent stage. Data of a sync block configuration appears from the framing circuit. The recording data is supplied to two rotary heads via a channel encoding circuit and a recording amplifier and recorded on a magnetic tape.

In the quantization circuit 7, the AC coefficient data is divided by the quantization step and the quotient is rounded to an integer. A
When the absolute value of the C coefficient data is C and the value of the quantization step is D, the rounded value Q (C) is expressed by the following equation. Q (C) = INT [{C + (D / 2)} / D]

Next, the activity detection performed by the classification circuit 8 and the classification based on it will be described. In this example, the classification is performed using the DC coefficient data and the absolute value of the AC coefficient data. The activity detection is to detect the fineness of the pattern of each macroblock. Visually, a macroblock with a fine pattern (high activity) has no noticeable distortion even if the quantization step is a little rough. On the other hand, if a block having a flat pattern (low activity) is roughly quantized, distortion tends to stand out. Therefore, it is effective to quantize macroblocks with high activity and coarse quantize macroblocks with low activity in units of macroblocks.

Further, in the case of a block with high brightness,
Even if the quantization is rough, the deterioration is not noticeable visually. Therefore, in this embodiment, the DC component of the DCT coefficient of the Y block is used to determine whether the brightness is high. When the value of the DC component of the Y signal is large, it is determined that the brightness is high. An example of the classification based on the brightness and activity is shown in FIG.

Four types of classes, class 0 to class 3, are prepared. Class 0 is the class in which coefficient data is quantized most finely, and quantization becomes coarser in the order of classes 1, 2, and 3. Therefore, class 3 quantization is the coarsest. As shown in FIG. 5, classification is first performed based on the DC coefficient of the Y signal. In this example, the DC coefficient is 9 bits and can take values from 0 to 511. As the threshold value, 400 is set, and when the DC coefficient is larger than 400, the class 3 is uniformly determined. That is, Y
A large DC coefficient of the signal means that the luminance is high, so that the class 3 having the coarsest quantization step is assigned to this case. Note that, for class 3 only, the values of all coefficient data are reduced to 1/2 by scaling as preprocessing for quantization.

When the DC coefficient is in the range of 0 to 400, the maximum absolute value of the AC coefficient of the DCT block,
That is, the class is determined based on the activity. For example, for the DCT block of the Y signal, the AC
When the maximum absolute value of the coefficient is in the range of 12 to 23, the DCT block is determined to be class 1. As another example, for the DCT block of the CR signal, the AC
When the maximum absolute value of the coefficient is in the range of 0 to 11, the DCT block is determined to be class 1. Figure 5
As can be seen from the above, the Y signal is compared with CR and CB,
The class is assigned a smaller number, and the CR signal, which is more visible, is finer in quantization than the CB signal.

The method of classifying by the DC coefficient is not limited to the above method, but it is also possible to use two threshold values and divide into three types of ranges. Also, instead of the maximum absolute value of the AC coefficient, the n-th power of the AC coefficient, A
The difference between the maximum and minimum values of the C coefficient may be used. Further, the number of coefficients larger than the threshold value may be used to determine the class. Furthermore, coefficient data used for activity detection may be limited to high frequency data.

The data amount estimator 9 can make the generated data amount of the buffering unit (5 macroblocks) equal to or less than the target value, and determines the quantization step of the smallest possible value. The estimator 9 determines the quantization number for every 5 macroblocks. Further, the area in the block is divided into eight, for example, and quantization is performed according to each area. Quantization considering the area will be described.

FIG. 6A shows an example of area division of coefficient data for a still block. Each number of 0 to 7 attached to each coefficient data represents an area number. The area number is defined such that the coefficient data is on the high frequency side as the area number increases. Area division is based on the fact that when the coefficient data is quantized, the higher the coefficient data, the less the quality of the restored image deteriorates even if the quantization is roughened. As can be seen from FIGS. 3 and 6, the area numbers change in ascending order according to the scanning (output) order of the coefficient data.

FIG. 6B is an example of area division of coefficient data for motion blocks. In the case of a motion block, the intra-field DCT is performed as described above.
The DCT coefficient generated in the first field is divided into areas as shown in the upper part of FIG. 6B, and the DC generated in the second field is divided.
The T coefficient is divided into areas as shown in the lower part of FIG. 6B. The area division may be the same between the luminance signal and the color signal, and the color signal may be shifted to the area number on the higher frequency side than the luminance signal.

FIG. 7 shows an example of the quantization table. In FIG. 7, 16 are identified by the quantization numbers QNo of 0 to 15.
A set of different quantization steps is provided. Each group is
The quantization step corresponds to each area number from 0 to 7. This quantization table has a change in which the quantization step becomes smaller as the quantization number QNo increases. In other words, if the quantization number QNo increases,
The quantization changes to fine ones. Since all quantization steps are expressed by powers of 2, a simple circuit can be used as a circuit for dividing coefficient data by these quantization steps.

Further, in this quantization table, quantization numbers 0 to 15 are adjusted in correspondence with classes 0, 1, 2, and 3. Between class 0 and class 1, and between class 1 and class 2, there is a difference of 3 in quantization number in the direction of increasing quantization step. Also, class 2
And the class 3, there is a difference of 2 in the quantization number in the direction of decreasing the quantization step. The class 3 quantization step should be the largest, but because of the initial scaling, as described above, there is an inversion relation.

In the data amount estimator 9 in FIG. 1, first, when detecting the data amount of 5 macroblocks, the generated data amount is quantized while referring to the quantization table of FIG. 7 provisionally as class 1. Detect by number. The optimum quantization number is selected from the 16 quantization numbers. That is, the quantization number is selected so that the generated data amount does not become larger than the target value and the quantization is most finely performed. This quantization number is supplied to the quantization number controller 10.

The quantization number controller 10 is supplied with the class information determined by the classifying circuit 8 based on the brightness and the absolute value of the AC coefficient, as described above. As class information, activity code A
T can be used. Since the quantization number is selected for every 5 macroblocks, the quantization step is fixed and the quantization step is adjusted according to the class. This adjustment is made using the quantization table of FIG.

The above embodiment is an example of a digital VTR for recording a digital video signal on a magnetic tape. However, the present invention can be applied to the case where a medium such as a disk other than a tape is used.

[0037]

According to the present invention, when the coefficient data generated by the cosine transform is quantized, the class is determined from both the brightness and the activity of the partial image, so that the image quality is not deteriorated. A high compression rate can be achieved.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment in which the present invention is applied to a recording data processing circuit of a digital VTR.

FIG. 2 is a schematic diagram used to describe a macroblock.

FIG. 3 is a schematic diagram showing an example of an output order of DCT coefficient data.

FIG. 4 is a schematic diagram showing an intra-field DCT process.

FIG. 5 is a schematic diagram used to explain classification according to an embodiment of the present invention.

FIG. 6 is a schematic diagram of an example of area regulation.

FIG. 7 is a schematic diagram of an example of a quantization table.

[Explanation of symbols]

 4 DCT circuit 5 Motion detection circuit 7 Quantization circuit 8 Classification circuit 9 Estimator

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Office reference number FI technical display location H03M 7/40 8842-5J H04N 1/41 B

Claims (3)

[Claims] 1. A digital image signal coding apparatus, wherein a digital image signal is coded by cosine transform, and AC coefficient data generated by coding is variable-length coded, in order to quantize the AC coefficient data. And a quantization table prepared in advance so as to select a quantization step in the quantizing means so that the data amount of the variable-length coded output in a predetermined period is kept below a target value. And a means for receiving the DC coefficient data and the AC coefficient data generated by the encoding, and determining a class indicating the fineness of the quantization of the partial image to be quantized. , For adjusting the quantization step of the predetermined period determined by the data amount estimating means according to the class of the partial image Encoding apparatus in a digital image signal comprising a stage. 2. The digital image signal coding apparatus according to claim 1, wherein the means for determining the class sets the quantization step to the maximum when the DC coefficient data is larger than a threshold value. An apparatus for encoding a digital image signal, wherein the class is fixed, and when the DC coefficient data is smaller than the threshold value, the class is determined based on the absolute value of the AC coefficient data. 3. The digital image signal encoding device according to claim 1, wherein the data amount estimating means is capable of suppressing the data amount of the variable-length encoded output to a target value or less for each of a plurality of blocks. An encoding device for a digital image signal, wherein the means for selecting a coding step and determining the class determines the class for each block.