KR100357089B1 - Frame reconstruction device of mpeg video encoder - Google Patents

Abstract

PURPOSE: A frame reconstruction device of an MPEG video encoder is provided to efficiently vary GOP(Group Of Picture) structure according to the quantity of motions and a variation in scenes in the event of encoding. CONSTITUTION: A frame reconstruction device of an MPEG video encoder includes a plurality of frame delay devices(30a,30b,30c,30d), a scene change detector(31), a prediction error analyzer(34), a controller(33), and a selector(32). The frame delay devices delay input video signals in unit of frame. The scene change detector compares a value obtained by accumulating a difference between the current scene and previous scene with a predetermined threshold to output a scene change generation signal. The prediction error analyzer compares the absolute value of the accumulated prediction error of the previous scene with the absolute value of the accumulated prediction error of the current scene and, when there is a difference between the two values larger than a predetermined threshold, outputs a control signal. The controller varies N value indicating frame distance between I pictures according to the scene change generation signal and changes M value indicating frame distance between I and P pictures according to the control signal output from the prediction error analyzer. The selector selectively outputs the video signals of the frame delay devices to rearrange the video signals according to the N and M values.

Description

Frame Reconstruction Device of MPEG Video Encoder

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an MPEG video encoder, and more particularly, to an apparatus for reconstructing a frame of an MPEG video encoder capable of efficiently varying a GOP structure according to a motion amount and a scene change during encoding.

Hereinafter, a conventional MPEG encoding system will be described with reference to the accompanying drawings.

FIG. 1 is a block diagram of a general MPEG video encoder, and FIG. 2 is a reconstruction block diagram of a general frame.

First, as illustrated in FIG. 1, the MPEG encoding system includes a DCT 3 for converting input image information in frame units into frequency components and a quantization for approximating the output signal of the DCT 3 to a finite number of values. A conversion unit 4, a VLC 5 for variable length coding the output signal of the quantization conversion unit 4, a buffer 6 for buffering and outputting the variable length coded image data, and the quantization conversion Inverse quantization converter 7 and inverse DCT converter 8 and the output signal of the inverse DCT converter 8 which inversely quantize and inverse DCT the image data to generate coefficient values for the difference image from the previous frame. The frame memory 10 stores and outputs the data in frame units, and the motion estimation unit 12 and the motion compensation unit 11 perform motion estimation and compensation on the output data of the frame memory 10.

Here, the reason for using the frame reconstructor 1 at the input of the encoder is as follows.

In general, in a digital imaging system, each frame of an image is sampled in pixel units. Since the image information configured as described above has a huge amount of information, an image compression (or encoding) technique is required for efficient transmission or storage.

This mainly uses a method of removing redundant information in the space and time domain.

The redundancy in the spatial domain is due to the small degree of change between adjacent pixels in one frame, and the redundancy in the time domain is due to the slight change in the movement of objects between adjacent frames.

Moving Picture Experts Group (MPEG) is an international standard for moving picture compression and reconstruction that compresses video signals by removing redundancy on the time and space axes.

Here, the redundancy of the time domain is eliminated through motion compensation. One screen encoding only an intra component according to the form of motion compensation, a P picture only forward prediction, and before and after It is divided into B screens that perform bidirectional prediction (forward and backward).

GOP is a collection of these I.P.B screens whose structure and characteristics can be represented by the values M and N.

(M value means the frame distance of the I or P screen, and N value indicates the size of the GOP and means the frame distance of one screen value.

For example, when (M, N) is (3,9), the picture order on the compressed bitstream is rearranged as shown in FIG. 2 (where N must be an integer multiple of M).

That is, the frame reconstructor is rearranged in order to more efficiently encode an image of a frame unit input by digging fixed values of M and N.

However, the frame reconstruction apparatus of the conventional MPEG video encoder has the following problems because it has a fixed GOP structure.

In other words, if the motion compensation is performed using a screen having a relatively long distance in a large motion image, the coding rate is lowered. Therefore, a relatively small M value must be used, but the GOP structure is fixed.

In addition, if a scene is changed to a P or B screen when a scene change occurs, more bits are generated than the bits allocated to the screen, resulting in deterioration of image quality. It is efficient to change the value and process it as an I screen, but it is difficult because the GOP structure is fixed.

The present invention has been made to solve the above problems of the conventional frame reconstruction apparatus of the MPEG video encoder, and it is possible to efficiently vary the GOP (GrouP Of Picture) structure according to the amount of motion and scene change during encoding. It is an object of the present invention to provide a frame reconstruction apparatus of an MPEG video encoder.

The frame reconstruction apparatus of the MPEG video encoder of the present invention for achieving the above object is a plurality of frame delay elements for delayed output of the input video signal in units of frames, a value accumulated by the difference between the current screen and the previous screen and the set threshold Compares the values and outputs the scene change occurrence signal, and compares the absolute value of the accumulated prediction error of the previous screen estimated with the macroblock unit with the absolute value of the accumulated prediction error of the current screen. A N error indicating a frame distance (size of the GOP) between the I screens is varied according to a prediction error analysis output unit for outputting an addition / deceleration control signal according to a prediction error analysis and a scene change generation signal of the scene change detection unit. Control means for varying the M value indicating the frame distance of 1, P screen by the control signal of the prediction error analysis output unit And a selection output unit for selectively outputting the video signal of the frame delay element by N and M values which are variably controlled by the control unit.

Hereinafter, a frame reconstruction apparatus of an MPEG video encoder of the present invention will be described in detail with reference to the accompanying drawings.

Figure 3 (a) is an upper block diagram of the frame reconstructor of the present invention, (b) is a detailed block diagram of the prediction error analysis output unit according to the present invention, (c) is an operation flowchart of the comparator of the prediction error analyzer , (d) is a detailed block diagram of a scene change detection unit according to the present invention.

First, the frame reconstruction apparatus of the MPEG video encoder of the present invention includes a plurality of frame delay elements 30a, 30b, 30c, 30d for delaying and outputting an input video signal in units of frames, and a previous picture value and a current picture value. Is a subtractor that compares and subtracts the macroblocks in units of macroblocks, an absolute value calculator 39, an accumulator 40, and the accumulator 40 that continue to add the difference value of the subtractor as an absolute value in units of frames. If the cumulative value of any one frame is larger than the set threshold value T ₂ , the scene change detection unit 31 including the comparator 42 for outputting the scene change generation signal and the predicted error value estimated in macroblock units are used. The absolute value of the absolute value calculation unit 35, the accumulator 36, and the register 38 that accumulate in the unit of frame on the basis of the absolute value, and the accumulated value of the current screen of the accumulator 36. Compare absolute values of prediction errors Between the threshold value (T ₁₎ at least once the difference consisting of the comparator 37 that outputs a subtraction control signal predictive error analysis Ex reokbu 34, the scene change detecting I picture by a transition occurrence signal of 31 Control means for varying an N value representing a frame distance (size of a GOP) and an M value representing a frame distance of an I and P screen according to the control signal of the prediction error analysis output unit 34. The frame delay elements 30a, 30b, 30c and 30d are controlled by the unit 33 and the N and M values that are variably controlled by the control means 33 (in the embodiment of the present invention, the M value is the maximum). And a selection output unit 32 for selectively outputting the video signals of the video signal to be rearranged.

The detailed operation of the comparator 37 of the predictive error analysis output unit 34 in the frame reconstruction apparatus of the MPEG video encoder of the present invention configured as described above is as follows.

First, the accumulated prediction error value (A) of the current screen is compared with the set threshold value (T ₁ ) + the prediction error value (B) of the previous screen (301). If the prediction error value (A) is large, the M value is decreased. (302).

When the prediction error value A is small, the accumulated prediction error value A of the current screen is compared with the set threshold value T1-the prediction error value B of the previous screen (303), and the prediction error value A is obtained. If large, M value is increased (304).

If the prediction error value A is small, the M value is not changed (305).

At this time, if the M value is larger than Mmax, it does not increase any more, and if the M value is smaller than Mmin, it does not decrease any more.

After performing steps 302, 304, and 305, the predicted error value A is stored in the register 38 (306).

4 is a diagram illustrating an output image of the selective output unit according to the change of the M value of the present invention, and illustrates a control operation of the control receiver 33.

For example, when Mmax = 3, an X ₄ signal is output if it is a B screen, and an X _4-M signal is output if it is a 1 or P screen.

FIG. 5 is a diagram illustrating an output image of a selective output unit according to a change in N value according to the present invention, and shows a GOP structure that changes according to a scene change generation signal, and (a) shows a scene when M = 3 and N = 6. There is no conversion to indicate a general arrangement.

However, if a scene change occurs between the 10th and 11th screens, the scene change detection unit 31 detects the scene change and detects the P in the sub GOP (M frame section, eg, PBB, IBB, etc.) where the scene change exists. If the screen exists, this P screen is switched to the I screen, and the I screen in the next GOP is switched to the P screen.

Therefore, the N value changes from 6 to 3 and back to 9.

In the frame reconstruction apparatus of the MFEG video encoder of the present invention as described above, the scene change generation signal is detected before the frame reconstruction, so it is used as a feed forward control signal, and the M control signal is added to the GOD currently encoded. This result is used as a feed back control signal because it is used in the next GOD encoding.

The frame reconstruction apparatus of the MPEG video coder of the present invention as described above can be effectively applied to a coding system having a high data rate because the GOD structure can be efficiently changed in response to a motion amount and a scene change. Therefore, it is effective to make encoding more efficient.

1 is a block diagram of a general MPEG video encoder

2 is a reconstruction block diagram of a general frame

3A is a block diagram of the frame reconstructor of the present invention.

(b) is a detailed block diagram of the predictive error analysis output unit according to the present invention;

(c) is a detailed block diagram of the scene change detection unit according to the present invention;

4 is a configuration diagram of an output image of the selective output unit according to the change of the M value of the present invention.

* Explanation of symbols for the main parts of the drawings

30a, 30b, 30c, 30d: frame delay element 31: scene change detection unit

32: selective output unit 33: control unit

34: prediction error analysis output unit 35, 39: absolute value calculation unit

36, 40: accumulator 37, 41: comparator

38: register

Claims (3)

A plurality of frame delay elements for delaying and outputting the input video signal in units of frames; A scene change detector for outputting a scene change generation signal by comparing a accumulated value of the difference between the current screen and the previous screen and a set threshold value; Predictive error analysis that outputs the control signal according to the predictive error analysis when the difference of more than the set threshold is compared by comparing the absolute value of the accumulated prediction error of the previous screen estimated in macroblock units with the absolute value of the accumulated prediction error of the current screen. An output unit, The N value indicating the frame distance (size of GOD) between the I screens is varied according to the scene change occurrence signal of the scene change detection unit, and M representing the frame distance of the I and P screens according to the control signal of the prediction error analysis output unit. A control means for varying the value; And a selective output unit for selectively outputting the video signal of the frame delay element by the variable controlled N and M values of the control unit. The method of claim 1, The scene change detection unit includes a subtractor for comparing and subtracting the previous screen value and the current screen value in macroblock units; An absolute value calculator and accumulator which continuously adds the difference value of the subtractor as an absolute value in units of frames; And a comparator for outputting a scene change generation signal when a cumulative value of any one frame of the accumulator is larger than a predetermined threshold value. The method of claim 1, The predictive error analysis output unit includes an absolute value calculator and an accumulator for adding the estimated error value estimated in macroblock units as absolute values in frame units. And a comparator configured to compare an absolute value of the accumulated prediction error of the previous register of the register with an absolute value of the accumulated prediction error of the current display of the accumulator and output an accel / decel control signal when a difference is greater than a predetermined threshold. Frame reconstruction apparatus of MPEG video encoder.