JPH10108184A - Image data processing unit and its method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the image data processing unit in which production of an overflow is effectively suppressed without deteriorating the image quality. SOLUTION: In quantization sections 202-1 to 202-n, a plurality of quantization indices including maximum quantization indices are used for quantization and a code length at the quantization level is obtained. An object code length decision section 206 obtains a final object code amount S206 to satisfy the object code length based on the code length above. Simultaneously in the case that the code length is longer than the object code length even when the maximum quantization index is in use, it is discriminated that an overflow takes place. A binary search section 208 obtains a quantization index S208 satisfying the final object code amount. A quantization section 3 uses the quantization index S208 to conduct quantization.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data processing apparatus and method for suppressing overflow due to an increase in bit rate in digital image compression processing.

[0002]

2. Description of the Related Art MPEG (Moving Pictures Expert Grou)
p) DCT (Discrete Cosine Transform) represented by the standard
In the image compression technique for digital video signals employing, code length control is generally performed so that a bit stream satisfies a desired rate. In this control, conventionally, the code length is feedback-controlled based on the relationship between the previous quantization step and the code length and the current average rate.

FIG. 7 is a configuration diagram of a conventional DCT type image compression apparatus 10. As shown in FIG. As illustrated in FIG. 7, the image compression device 10 includes a macroblock unit 1, a DCT unit 2, a quantization unit 3, a VLC unit 4, a buffer 5, and a quantization control unit 6.

In the image compression apparatus 10, an image corresponding to an input video signal is converted from a field format to a frame format, and then divided into macroblocks consisting of several DCT blocks in a macro flocking unit 1. You. The macro block S1 is output to the DCT unit 2. In MPEG, a 16 × 16 luminance signal (Y) and a color difference signal (Cb, Cr) overlap to form a macroblock. Specifically, in the case of 4: 2: 0, the macroblock includes four 8 × 8 blocks for Y and C
It is composed of a total of six (DCT) blocks including one 8 × 8 block for each of b and Cr.

The DCT unit 2 performs DCT (Discrete Cosine Transform) for each (DCT) block constituting the macroblock S 1, and outputs a DCT coefficient S 2 to the quantization unit 3. The quantization control unit 6 obtains a quantization index in consideration of the bit rate of the output S5 of the buffer 5, and outputs the quantization index S6 to the quantization unit 3. In the code length control in the MPEG2 test model generally used in moving image compression, quantization is controlled for each macroblock based on an activity (visibility of image quality deterioration with respect to the macroblock), and a virtual buffer of the virtual buffer is controlled. Remaining amount,
Feedback control is performed using the relationship between the quantization index and the generated code length in the previous encoding. In this control, feedback is performed so as to keep the average bit rate constant.

[0006] The quantization unit 3 quantizes the DCT coefficient S2 based on the quantization index S6, and outputs a quantization level S3 as quantized data to the VLC unit 4.
The VLC unit 4 sets the quantization level S3 to VLC (Variable Le
ngth Code (variable length code), and outputs the bit stream S4 to the buffer 5. At this time, the relationship between the generated code length and the quantization index is fed back to the quantization control unit 6 via the buffer 5. Bit stream S4
Is output as a bit stream S5 via the buffer 5.

[0007]

However, in the conventional image compression apparatus 10 described above, since the quantization control unit 6 performs feedback control so as to keep the average bit rate constant, a predetermined number of images can be obtained. It is difficult to control the frame to be within a predetermined bit rate. Therefore, when recording on a recording medium such as a tape,
The length of the frame unit is undefined. As a result, it becomes difficult to identify a break in an image on a recording medium, which is extremely inconvenient in operability such as shuttle playback of editing. In addition, in places such as scene changes, past statistical data cannot be used, and the rate fluctuates instantaneously, depending on the application, an overflow occurs, part of the DCT coefficient data is lost, and extreme image quality deterioration is caused. Could be

Also, because of the feedback control, when the damping is reduced, the response is quick but vibratory, and when the damping is increased, the vibration is reduced but the response is delayed. As described above, the conventional image compression device 10 is
It is not suitable for such applications.
In particular, the above-mentioned overflow problem is fatal.
That is, if it is known in advance that an overflow will occur, the higher-order coefficients can be discarded with priority given to the more important lower-order coefficients, but for that purpose, it is necessary to predict the presence or absence of overflow.

In order to solve such a problem of the overflow, there has been proposed a method of performing control such that recording is performed at a fixed length rate by feedforward control. However, in this method, since a very large quantization step is used to reduce the frequency of occurrence of overflow, a new problem that the image quality is deteriorated as a whole arises.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems of the prior art, and has as its object to provide an image data processing apparatus and an image data processing apparatus capable of effectively suppressing the occurrence of overflow without deteriorating the image quality.

[0011]

To achieve the above object, an image data processing apparatus according to the present invention determines a quantization index so that encoded data of digital image data is equal to or less than a target code length, An image data processing device for quantizing the digital image data based on the quantization index, comprising: a transform unit for orthogonally transforming the digital image data into a frequency domain; Quantizes the orthogonally transformed digital image data using the quantization index of, calculates a code length from each of the quantization levels, and determines a final target code length based on the code length and the target code length. Code length determining means, and the coded data quantized using the maximum quantization index is equal to or less than the target code length. An overflow judging means for judging an overflow if not, and a binary search method for finding a minimum quantization index at which the coded data becomes equal to or less than the determined final target code length when the overflow is not judged. And quantization means for quantizing the orthogonally transformed digital image data based on the determined quantization index.

In the image data processing apparatus according to the present invention, the overflow judging means judges that an overflow has occurred if the encoded data quantized using the maximum quantization index does not fall below the target code length. As a result, before performing the binary search process, the overflow can be appropriately detected, and the incorrect search of the quantization index in the binary search means can be appropriately avoided.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image data processing apparatus according to an embodiment of the present invention will be described below. This image data processing apparatus is used, for example, in a digital VTR that records a digital video signal on a magnetic tape by a rotating head.

FIG. 1 is a configuration diagram of a DCT type image compression apparatus 100 according to an embodiment of the present invention. FIG. 6 is a timing chart of processing in the image compression apparatus 100 shown in FIG. In the following description, processing is performed so as to satisfy a desired code length in units of one frame.

[0015] As shown in FIG.
Are the macroblocking unit 1, the DCT unit 2, the quantization unit 3,
VLC unit 4, buffer 5, activity detection unit 20
1, quantization units 202-1 to 202-n, encoding unit 203
-1 to 203-n, integrating units 204-1 to 204-n, F
IFO sections 205-1 to 205-n, target code length determination section 2
06, a FIFO unit 207 and a binary search unit 208.

In the image compression apparatus 100, an image corresponding to an input video signal is converted from a field format to a frame format, and then divided into macro blocks composed of several DCT blocks in a macro flocking unit 1. You. The macro block S1 is output to the DCT unit 2. In MPEG, a 16 × 16 luminance signal (Y) and a color difference signal (Cb, Cr) overlap to form a macroblock. Specifically, in the case of 4: 2: 0, the macroblock includes four 8 × 8 blocks for Y and C
It is composed of a total of six (DCT) blocks including one 8 × 8 block for each of b and Cr.

The DCT unit 2 performs DCT (Discrete Cosine Transform) for each (DCT) block constituting the macroblock S1, and converts the DCT coefficient S2 shown in FIG.
02-1 to 202-n. The activity detection unit 201 detects the activity of the macroblock S1, and detects the detected activity S shown in FIG.
201 is output to each of the quantization units 202-1 to 202-n.

The quantization units 202-1 to 202-n quantize the DCT coefficient S2 at mutually different quantization steps,
Quantization levels S202-1 to S202-1 shown in FIG.
−n to encoders 203-1 to 203-n. At this time, the quantization units 202-1 to 202-n
Activity S2 from activity detector 201
01, if the image quality deterioration is likely to be conspicuous, the quantization step of each quantization unit is made finer.If the image quality deterioration is not conspicuous, the quantization step of each quantization unit is made coarse to perform quantization. Do.

The encoding units 203-1 to 203-n respectively convert the quantization levels S202-1 to S202-n into VLCs.
(Variable Length Code) is obtained, and the code lengths S203-1 to S203-1 shown in FIG.
S203-n is calculated by integrating units 204-1 to 204-n, respectively.
And FIFO units 205-1 to 205-n.
The encoding units 203-1 to 203-n may obtain the code length when VLC is performed without actually performing VLC.

The integrating units 204-1 to 204-n integrate the code lengths S203-1 to S203-n for one frame, respectively, to obtain an integrated value shown in FIG. As shown in FIG. 6F, the FIFO units 205-1 to 205-n
The code lengths S203-1 to S203-n are delayed by the time of integration in the integration units 204-1 to 204-n,
The code lengths are output as FIFO lengths S205-1 to S205-n.

The target code length determining unit 206 includes an integrating unit 204
Integration results from S2-1 to S204-n S204-1 to S20
Based on 4-n, a final target code length for each macroblock is determined as shown below.

For example, in FIG. 1, the total number of quantization steps is 32 from 0 to 31, and n = 4, and the quantization unit 202-j (1 ≦ j ≦ n, j is an integer) The conversion index q [j] is determined as shown in Table 1 below.

[Table 1]

Here, the maximum quantization index (= 3)
1) is included in the quantization index q [j]. Assuming that the code length of the i-th macroblock by the quantization unit 202-j is 11 (i, q [j]), the quantization unit 202-j
j, the integrated value of the code length for one frame Σ _i ll (i,
q [j]) is plotted, for example, as shown in FIG.
As shown in FIG. 2, quantization that satisfies the target code length (tgt) is performed by combining the code length by the quantization of the quantization unit 202-2 with the code length by the quantization of the quantization unit 202-3. You can see that it is in between. Therefore, by linear approximation of these two points, a final target code length for each macroblock that satisfies the target code length (tgt) is obtained. Assuming that the final target code length of the macroblock i is 11 ＾ (i), the target code length determination unit 206 obtains the final target code length according to the following equation (1).

[Number 1] ll ^ (i) = [{Σ _k ll _(k, q [2]) - tgt} · ll (i , q _{[3]) + {tgt-Σ k ll} (k, q [3] )} · ll (i, q [2])] _{/ {Σ k ll (k,} q _{[2]) - Σ k ll (k,} q [3])} ... (1) in the target code length determining part 206 Figure 6 (G) determined
The final target code length of the macroblock i shown in FIG.
Is output to the binary search unit 208.

In the case shown in FIG. 3, if the quantization index that satisfies the target code length (tgt) is smaller than the quantization index of the quantization unit 202-1, the quantization is performed. The target code length (tgt) is obtained by multiplying the code length using the quantization index of the unit 202-1 by a constant.
Is obtained for each macroblock that satisfies That is, the final target code length of the macroblock i is l
Assuming that l） (i), it is obtained by the following equation (2). Here, the quantization index of the quantization unit 202-1 is the smallest quantization index among the quantization indexes provided in the target code length determination unit 206.

[0025]

[Number 2] ll ^ (i) = _{[tgt / {Σ k ll (k} , q [1])}] · ll (i, q [1]) ... (2) Further, in the case of Figure 4, Although the final target code length is calculated from the above equation (2), overflow occurs because the code length is long even in the quantization unit 202-4 that takes the maximum quantization index. In such a case, the target code length determining unit 206 outputs to the binary search unit 208 the final target code length S206 including data with a flag indicating overflow. Further, the target code length determining unit 206 may output only the data indicating the overflow to the binary search unit 208.

The FIFO unit 207 receives the D from the DCT unit 2
The DCT coefficient S2 is converted to the DCT coefficient S20 so that the CT coefficient S2 is output to the binary search section 208 at the same timing as the target code length 11 ^ (i) from the target code length determination section 206.
7 is output to the binary search unit 208.

The binary search section 208 determines the minimum quantization index for which the code length of the macroblock i satisfies the final target code length 11 長 (i) determined by the target code length determination section 206. .

Hereinafter, the binary search unit 2 will be described with reference to FIG.
The process in 08 will be described. FIG. 5 shows 32 quantization indexes and plots of the code lengths when the macroblock i is quantized and VLC is performed using the quantization indexes. Here, the minimum q that satisfies the final target code length 11 ＾ (i), that is, q _i = min _j (11 [i, qj] ≦ l
l ＾ (i)). Although the binary search method requires monotonicity of the function, the code length increases as the quantization decreases, so that the generality is not lost even in the example shown in FIG. The binary search method is realized by the binary search unit 208 in the following steps.

Step 1: If the overflow flag included in the final target code length S206 from the target code length determination unit 206 is not set, it is understood that the solution exists in the range from q0 to q31. Therefore, a code length at q15, which divides the range of the solution into two, is obtained. Then, ll
Since (i, q15)> ll ＾ (i), the solution exists in the range from q16 to q31.

Step 2: Since it is known that the solution exists between q16 and q31, a point which divides the range of the solution into two and a code length at q23 are obtained. Then, ll
Since (i, q23) ≦ 11 ＾ (i), the solution exists in the range from q16 to q23.

Step 3: Since it is known that the solution exists between q16 and q23, the code length at q19, which divides the range of the solution into two, is determined. Then, ll
(I, q19) ≦ 11 ＾ (i) (actually, 11 (i, q
19) = 11 ＾ (i)), the range of existence of the solution is q
16 to q19.

Step 4: Since it is known that the solution exists between q16 and q19, the code length at q17, which divides the range of the solution into two, is determined. Then, ll
Since (i, q17)> 11 ＾ (i), the existence range of the solution exists between q18 and q19.

Step 5: Since it is known that the solution exists in the range from q18 to q19, a point which divides the range of the solution into two and the code length at q18 are obtained. Then, ll
(I, q18) ≦ 11 ＾ (i) (actually, 11 (i, q
18) = 11 ＾ (i)), the existence range of the solution is q
18 to q18, that is, q18.

As described above, in the binary search, always "log
Existence range of _two solutions "can be obtained. Further, since only one quantization is checked in each step, it can be sufficiently realized in hardware. However, in the processing of steps 1 to 5, the maximum quantization index is not checked. This is because it is assumed that the result of the largest quantization index is smaller than the final target code length. However, actually, even if the maximum quantization index is larger than the final target code length, overflow occurs. As described above, in the image compression device 100, the quantization unit 202-n also checks the maximum quantization index, and the result is indicated by the overflow flag included in the final target code length S206. as a result,
The occurrence of overflow can be easily detected, and the binary search unit 2
It is possible to avoid selecting an incorrect quantization index at 08.

The quantization unit 3 performs quantization based on the quantization index S208 determined by the binary search unit 208, and sets the quantization level S3 shown in FIG.
Output to section 4. The VLC unit 4 performs variable length coding (VLC) on the quantization level S3 to generate a bit stream S4 shown in FIG. At this time, if overflow occurs, processing such as protection of the low-frequency DCT coefficient is performed. The generated bit stream S4 is output as a bit stream S5 via the buffer 5 at a predetermined output rate.

As described above, the image compression apparatus 100
According to this, the code length of one frame can be transmitted at a determined bit rate, and the code length can be controlled without being affected by a scene change due to overflow detection. Further, realization by hardware is sufficiently possible. More specifically, according to the image compression apparatus 100, while taking into account the local properties of the digital image signal, even in an application such as a VTR, image breakdown due to an overflow detection error is prevented, and code length control of the feedforward system is performed. Can be realized effectively.

The present invention is not limited to the above embodiment. For example, in the above-described embodiment, DCT is exemplified as transform coding, but wavelet transform, Haar transform, KL transform, or the like may be used as transform coding.

In the above embodiment, the present invention is applied to the case where a digital video signal is recorded on a VTR. However, the recording medium does not need to be a tape, but may be a magneto-optical recording disk or a hard disk. Also, the present invention
The present invention may be applied to a device that does not use a recording medium, such as a communication system.

In the above embodiment, 4: 2: 0
Although the macro block structure of the format is used, the present invention may use a macro block structure such as 4: 2: 2, 4: 4: 4, and 4: 1: 1. Further, the number of DCT blocks constituting the macro block is also arbitrary.

Further, in the above-described embodiment, the control is performed such that the bit rate is stored in units of one frame. However, the control may be performed such that the bit rate is stored in units larger or smaller than this.

Further, in the above-described embodiment, reduction of a still image has been described.
The CT unit 102 may perform MC (Motion Compensation: motion compensation) and DCT so as to apply to reduction of a moving image.

Further, in the above-described embodiment, when the activity is obtained, the data before the DCT is used. However, the data after the DCT may be used. In the above-described embodiment, the binary search is solved in the range of all the quantization indices. However, the present invention solves the binary search in the range of some of the quantization indices. Is also good.

In the above-described embodiment, the quantization unit 2
As a method of predicting the final target code length for each macroblock from the code lengths of 02-1 to 202-n, interpolation using a linear approximation method was used, but an approximation method using a higher-order function using more points was adopted. You may.

Further, in the above-described embodiment, the final target code length of each macro block is individually allocated. However, the present invention allocates the final target code length equally to all macro blocks. You may do it.

[0045]

As described above, according to the image data processing apparatus and method of the present invention, a digital video signal can be transmitted at a predetermined bit rate.
Moreover, code length control that is not affected by scene changes or the like can be performed by overflow detection. Further, realization by hardware is sufficiently possible.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a DCT type image compression apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram in which an integrated value Σ _i ll (i, q [j]) of a code length for one frame by a quantization unit is plotted.

FIG. 3 is a diagram illustrating another example in which the integrated value 符号_i ll (i, q [j]) of the code length for one frame by the quantization unit is plotted.

FIG. 4 is a diagram illustrating still another example in which the integrated value Σ _i ll (i, q [j]) of the code length for one frame by the quantization unit is plotted.

FIG. 5 is a diagram for explaining a binary search method.

FIG. 6 is a timing chart of processing in the image compression device shown in FIG. 1;

FIG. 7 is a configuration diagram of a conventional DCT type image compression apparatus.

[Explanation of symbols]

10, 100 image compression device, 1 macroblock conversion unit, 2 DCT unit, 3 quantization unit, 4 VLC unit, 5 ...
Buffer, 201... Activity detector, 202-1
.. 202-n... Quantizers, 203-1 to 203-n... Encoders, 204-1 to 204-n.
205-n: FIFO unit, 206: target code length determination unit.

Claims (16)

[Claims] An image data processing apparatus which determines a quantization index so that encoded data of digital image data is equal to or less than a target code length, and quantizes the digital image data based on the quantization index. Transform means for orthogonally transforming the digital image data into the frequency domain, and quantizing the orthogonally transformed digital image data using a plurality of mutually different quantization indices including a maximum quantization index; Target code length determining means for determining a final target code length based on the code length and the target code length, and coded data quantized by using the maximum quantization index. An overflow judging means for judging an overflow when the code length is not shorter than the code length; When the overflow is not determined, the quantization index determining means for determining a minimum quantization index by which the encoded data is equal to or less than the determined final target code length by a binary search method; and the determined quantization index. And a quantizing means for quantizing the orthogonally transformed digital image data based on the image data. 2. The method according to claim 1, wherein said target code length determining means uses a linear approximation method when said target code length exists between two code lengths obtained using different quantization indices. The image data processing device according to claim 1, wherein the code length is determined. 3. The target code length determining means, when the target code length is longer than a code length obtained by using the minimum quantization index provided in the target code length determining means, the minimum quantization index. 2. The image data processing apparatus according to claim 1, wherein a value obtained by multiplying the code length obtained by using a constant by a constant is used as the final target code length. 4. A data dividing means for dividing the digital image data into a plurality of block data, wherein the converting means, the target code length determining means, the overflow determining means, the quantization index determining means, and the quantum The image data processing apparatus according to claim 1, wherein the conversion unit performs each processing in units of the block data. 5. An apparatus according to claim 1, further comprising: activity detecting means for detecting an activity based on the digital image data or the orthogonally transformed digital image data, wherein the target code length determining means is configured to detect an activity based on the detected activity. 2. The image data processing apparatus according to claim 1, wherein the quantization is performed. 6. The image data processing apparatus according to claim 1, wherein said target code length determining means determines a final target code length for each frame. 7. The image data processing apparatus according to claim 1, further comprising a variable length coding unit that performs variable length coding on the quantized quantization level. 8. The image data processing apparatus according to claim 1, wherein said conversion means performs discrete cosine transform of the digital image data. 9. An image data processing method for determining a quantization index so that encoded data of digital image data is equal to or less than a target code length, and quantizing the digital image data based on the quantization index. The digital image data is orthogonally transformed into the frequency domain, the orthogonally transformed digital image data is quantized using a plurality of mutually different quantization indices including a maximum quantization index, and codes are respectively encoded from their quantization levels. Determine the final target code length based on the code length and the target code length.If the encoded data quantized using the maximum quantization index does not fall below the target code length, an overflow occurs. Is determined, and when the overflow is not determined, the encoded data is Image data processing for obtaining a minimum quantization index that is equal to or less than the determined final target code length by a binary search method, and for quantizing the orthogonally transformed digital image data based on the determined quantization index. Method. 10. When the target code length is between two code lengths obtained using different quantization indexes, the final target code length is determined using a linear approximation method. 2. The image data processing method according to 1. 11. When the target code length is longer than the code length obtained by using the minimum quantization index provided in the target code length determination processing, the target code length is obtained by using the minimum quantization index. The image data processing method according to claim 9, wherein a value obtained by multiplying a code length by a constant is used as the final target code length. 12. The digital image data is divided into a plurality of block data, and the conversion process, the target code length determination process, the overflow determination process, the quantization index determination process, and the quantization process are performed by the block process. 10. The image data processing method according to claim 9, wherein the method is performed in data units. 13. A method for detecting an activity based on the digital image data or the orthogonally transformed digital image data, and performing the quantization based on the detected activity in the target code length determination processing. Item 10. The image data processing method according to Item 9. 14. The image data processing method according to claim 9, wherein the final target code length is determined for each frame. 15. The image data processing method according to claim 9, wherein said quantized quantization level is variable-length coded. 16. The image data processing method according to claim 9, wherein said orthogonal transform is a discrete cosine transform.