KR20100016689A - Fast inter mode decision method in h.264 encoding - Google Patents

Abstract

PURPOSE: A fast inter mode decision method in h.264 encoding is provided to determine an inter mode with less calculating load without propagating an error in selection of a wrong mode. CONSTITUTION: Among a motion vector between a current macro block and the first reference frame and a motion vector between the current macro block and the second frame, whether at least one is a zero motion vector is determined(502). The current macro block is compared with the first correspondence macro block and the second correspondence macro block and non-varying or not is determined(504). The current macro block is encoded by a direct mode(506).

Description

H.264 Fast Inter Mode Decision Method in H.264 Encoding

The present invention relates to H.264 encoding, and more particularly, to a method of determining an inter mode at high speed in H.264 encoding.

 Video coding standards such as MPEG-1, MPEG-2, MPEG-4, H.263, and H.264 video standard coding methods have been developed and play a leading role in digital media. This compression technology is making a tremendous change in the field of VCD, DVD, and digital TV. Among them, H.264 compression technology is a video compression method used for mobile reception DMB and has high compression performance. It is used in the field.

H.264 is the latest video standard project of the ITU-T Video Coding Experts Group (VCEG) and ISO / IEC Moving Picture Experts Group (MPEG), and is interactive or non-interactive with improved compression performance over previous compression technologies. The application provides friendly video footage on the specified network.

In H.264, a block having 16 * 16 pixels is formed into one macro block, divided into blocks of various sizes, and intra prediction in the pixel region for more accurate motion prediction and compensation, Macroblock mode is determined by various block sizes, multiple reference techniques, and rate-distortion optimization techniques, which can save 50% or more bits at the same image quality compared to existing video compression standards.

1 is a diagram illustrating a configuration of a general H.264 encoder.

Referring to FIG. 1, a general H.264 encoder includes a coding controller 110, a transform and quantizer 120, an entropy encoder 130, an inverse transform and inverse quantizer 140, and a deblocking filter 150. The image storage unit 160, a motion estimator 170, a motion compensator 180, and an intra predictor 190 are included.

When the input sequence is an intra frame, the prediction sequence is performed by the intra predictor 190 through the encoding controller 110. Subsequently, the predicted video signal is transformed and quantized by the transform and quantizer 120, and is output through the entropy encoder 130, while passing through the inverse transform and inverse quantizer 140 and the deblocking filter 150. It is stored in the reference image storage unit 160 to be used in the prediction process of the next frame.

Meanwhile, when the input sequence is an inter frame, the motion estimator 170 detects a motion vector for the input sequence and determines an optimal macro block. Then, the motion compensator 180 stores the reference image storage unit ( A prediction frame is generated using the reference image previously stored in 160. The predicted inter frame is integer transformed and quantized by the transform and quantizer 120, and is output through the entropy encoder 130, while passing through the inverse transform and inverse quantizer 140 and the deblocking filter 150. It is stored in the reference image storage unit 160 to be used in the prediction process of the next frame.

2 is a diagram for describing a macroblock type of an inter mode for H.264 encoding.

As shown in Fig. 2, the macro block type of the inter mode is divided into 16 * 16, 8 * 16, 16 * 8 and 8 * 8 modes, and the 8 * 8 mode is again P8 * 8, P4 * 8 and P8. It is divided into * 4, P8 * 8 and P4 * 4 modes. In the inter mode, all motion estimation processes are performed for each block type.

3 and 4 illustrate a sub mode of an intra mode macro block for H.264 encoding, and FIG. 3 illustrates a sub mode of a 4 * 4 block type, and FIG. 4 illustrates a 16 * 16 block type. The sub mode is shown.

Intra prediction is vertical, horizontal, diagonal down left, diagonal down right, vertical right, and horizontal down for intra 4 * 4 block types. Intra prediction is performed on a total of nine modes including a prediction mode with reference to pixel values in down, vertical left, and horizontal up directions, and a DC prediction mode. In the case of the block type, prediction is performed for four cases of vertical, horizontal, DC, and planar modes.

When inter or intra prediction is performed, a rate-distortion (RD) value is calculated and an RD cost is calculated to select an optimal mode to select an optimal mode for the macroblock.

Several algorithms have been proposed to determine the intermode at high speed. The approach for fast decision of inter mode can be classified into two types. The first method is to reduce candidate mode by using preset constraints in inter mode prediction, and the second method is to skip skip mode of macro block. It is predicting early. This approach is based on the fact that in H.264 coding, the skip mode is dominant over other candidate modes.

However, both approaches have had problems. For the first approach, the efficiency is determined by the search method. In general, the first approach uses a full search method, and when using the full search method, motion estimation must be performed for all modes. Inefficient in terms of complexity.

In addition, the second method has a problem that such an error is continuously propagated when an inappropriate mode is selected.

In the present invention, to solve the problems of the prior art as described above, it is proposed a method that can determine the inter mode when encoding H.264 at high speed.

Another object of the present invention is to propose a method for determining an inter mode with a small amount of calculation without propagating an error when selecting a wrong mode.

Other objects of the present invention may be derived by those skilled in the art through the following examples.

In order to achieve the above object, according to an aspect of the present invention, there is provided a method for determining an inter mode in H.264 encoding, comprising: determining whether a current frame is a B-picture; If the current frame is a B-picture, at least one of a motion vector of the current macroblock and the first reference frame of the current frame and a motion vector of the current macroblock and the second reference frame is a zero motion vector. Determining whether or not (b); When at least one of the motion vector of the current macroblock and the first reference frame and the motion vector of the current macroblock and the second reference frame is a zero motion vector, the current macroblock and the current macroblock It is determined whether a first corresponding macroblock of the corresponding first reference frame is invariable and whether the current macroblock and the second corresponding macroblock of the second reference frame corresponding to the current macroblock are invariant. Step (c); And (d) encoding the current macroblock by direct mode when the current macroblock is invariant compared to the first and second corresponding macroblocks. A fast inter mode decision method is provided.

The motion vector of step (b) is a direct motion vector.

The first corresponding macro block and the second corresponding macro block of step (c) are macro blocks at the same position as the current macro block in the first reference frame and the second reference frame.

In step (c), it is determined whether the current macroblock is invariant with the first corresponding macroblock and the second corresponding macroblock by using a sum of absolute difference (SAD) scheme.

According to another aspect of the invention, determining whether the current frame is a B-picture (a); (B) determining whether at least one of the direct motion vectors of the current macro block of the current frame is a zero motion vector; (C) determining non-variability between the current macroblock and the corresponding macroblock of the reference frame when the condition of step (b) is satisfied; And encoding (d) by the direct mode when the current macro block and the corresponding macro block are non-variable.

The present invention has the advantage that the inter mode can be determined with a small amount of calculation without error propagation when selecting the wrong mode.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the fast inter mode determination method in H.264 encoding according to the present invention.

5 is a flowchart illustrating the overall flow of a method for determining a fast inter mode in H.264 encoding according to the present invention.

Referring to FIG. 5, it is first determined whether a current frame to be coded is a B picture (step 500). The present invention is applied to B-pictures among I-pictures, B-pictures, and P-pictures in H.264 coding. The reason why the method proposed in the present invention is limited to the B-picture is to not propagate an error due to wrong mode selection. The error is not propagated because the B-picture is not used as a reference frame of another frame.

If the current frame to be coded is not a B-picture, a general inter mode prediction process is performed (step 508). If the current frame to be coded is a B-picture, a fast inter mode determination method after step 502 is performed.

If the current frame to be coded is a B-picture, at least one of a motion vector of the macro block to be currently coded and the first reference frame List0 and the second reference frame List1 is a zero motion vector. Determine (step 502).

The B-picture is coded using two reference frames of the first reference frame List0 and the second reference frame List1, and in step 502, motion vectors with two reference frames are obtained.

If at least one of the motion vectors with the two reference frames is a zero motion vector, the process proceeds to step 504, otherwise a general inter mode prediction scheme is performed (step 508).

According to a preferred embodiment of the present invention, the motion vector with the first reference frame and the second reference frame considered in step 502 is a direct motion vector. The direct motion vector of the current macro block and the first reference frame and the second reference frame does not require a separate operation for determining in step 502, which can be obtained using the motion vectors of previously coded macro blocks.

If at least one of the direct motion vectors of the current macro block and the first and second reference frames is a zero motion vector, whether the corresponding macro blocks of the current macro block and the first reference frame and the second reference frame are stationary. It is determined whether or not (step 504).

Here, the corresponding macro block means a macro block of the reference frame existing at the same position as the current macro block of the current frame.

6 is a diagram illustrating a relationship between a current macro block and a corresponding macro block.

Referring to FIG. 6, the first corresponding macro block is a macro block at the same position as the current macro block in the first reference frame List0, and the second corresponding macro block is the same as the current macro block in the second reference frame List1. Macro block of position.

When the corresponding macroblock of the first reference frame and the second reference frame is found, it is determined whether the current macroblock and the first corresponding macroblock are invariable and whether the current macroblock and the second corresponding macroblock are invariant.

Whether the current macroblock and the corresponding macroblock are invariant is to determine whether the pixel values of the current macroblock and the corresponding macroblock are substantially the same.

According to an embodiment of the present invention, a Sum of Absolute Difference (SAD) scheme may be used to determine whether the current macroblock and the searched corresponding macroblock are invariant. That is, if the difference between the pixel values of the current macroblock and the corresponding macroblock is less than or equal to a preset threshold, it is determined that the current macroblock and the corresponding macroblock are non-variable, otherwise the current macroblock and the corresponding macroblock are variable. To judge.

Of course, the determination of whether or not invariant is not limited to the SAD method, it will be apparent to those skilled in the art that various calculation methods may be used.

If the first corresponding macro block and the second corresponding macro block are not invariant compared to the current macro block, a general inter mode prediction process is performed (step 508).

In FIG. 5, the determination process of step 504 is performed after the determination process of step 502. However, the order of the determination process may be changed. If the conditions of step 502 and the conditions of step 504 are satisfied, step 506 is performed.

If both the conditions of step 502 and step 504 are satisfied, various modes existing in the inter mode are omitted and inter mode coding is performed by the direct mode (step 506).

The direct mode is a mode in which a motion vector can be derived through one of a spatial technique and a temporal technique in a decoder, and the bit rate is significantly reduced because it does not require motion information such as a reference frame index and a motion vector.

The spatial direct mode is a method of spatially predicting and obtaining a first and second reference frame index and a motion vector from neighboring macroblocks of a current macroblock to be coded, and the temporal direct mode is a motion vector of a corresponding macroblock present in a reference frame. By using the information, two direct motion vectors of the macro block to be currently coded are obtained.

That is, the present invention performs coding by the direct mode without calculating the rate-distortion cost by coding the various inter modes one by one when the above two conditions are satisfied, which can greatly reduce the amount of computation required for inter mode coding. There is an advantage.

7 is a table showing the PSNR difference (PSNR), the bit rate difference (Bits (%)) and the time difference (Time (%)) when general inter mode coding and the present invention are applied.

Figure 7 shows the results for the image of the QCIF standard of the image size of 176 * 144, the results of the experiment on three images (image A, image B).

Experiments were conducted for various QP (Quantization Parameter) conditions of 20, 24, 28, 32, and 34. The experiments were performed for the case of the image having the IBP structure and the case of the IBBP structure.

Referring to FIG. 7, it can be seen that the time required for coding over 59% when the IBP structure is reduced in the case of the image A and the time required for over 71% when the IBBP structure is reduced. In addition, in case of the image B, the time required for the IBP structure is reduced by 51% or more, and the time required for the IBBP structure is reduced by 59% or more.

It can be seen that the change in PSNR and bit rate is not large compared to the reduction in coding time. In addition, it can be seen that the present invention is applied to the B-picture, and thus, the time required for the IBBP structure can be further reduced.

 Although the above has been described with reference to a preferred embodiment of the present invention, those skilled in the art will be variously modified and modified within the scope of the present invention without departing from the spirit and scope of the present invention described in the claims below. It will be appreciated that it can be changed.

1 is a diagram illustrating a configuration of a general H.264 encoder.

2 is a diagram for explaining a macroblock type of an inter mode for H.264 encoding.

3 and 4 illustrate a sub mode of an intra mode macro block for H.264 encoding.

5 is a flowchart illustrating the overall flow of a method for determining a fast inter mode in H.264 encoding according to the present invention.

6 is a diagram showing a relationship between a current macro block and a corresponding macro block.

7 is a table showing the PSNR difference (PSNR), the bit rate difference (Bits (%)) and the time difference (Time (%)) when general inter mode coding and the present invention are applied.

Claims (8)

In the method of determining the inter mode in H.264 encoding, Determining whether the current frame is a B-picture (a); If the current frame is a B-picture, at least one of a motion vector of the current macroblock and the first reference frame of the current frame and a motion vector of the current macroblock and the second reference frame is a zero motion vector. Determining whether or not (b); When at least one of the motion vector of the current macroblock and the first reference frame and the motion vector of the current macroblock and the second reference frame is a zero motion vector, the current macroblock and the current macroblock It is determined whether a first corresponding macroblock of the corresponding first reference frame is invariable and whether the current macroblock and the second corresponding macroblock of the second reference frame corresponding to the current macroblock are invariant. Step (c); And And (d) encoding the current macroblock by direct mode when the current macroblock is invariant compared to the first and second corresponding macroblocks. How to determine fast inter mode in encoding. The method of claim 1, The method of claim 1, wherein the motion vector of step (b) is a direct motion vector. The method of claim 1, The first corresponding macro block and the second corresponding macro block of the step (c) are macro blocks at the same position as the current macro block in the first reference frame and the second reference frame. How to determine intra mode. The method of claim 1, In the step (c), it is determined whether the current macroblock is invariant with the first corresponding macroblock and the second corresponding macroblock by using a Sum of Absolute Difference (SAD) method. Fast Intra Mode Determination Method in H.264 Encoding. Determining whether the current frame is a B-picture (a); (B) determining whether at least one of the direct motion vectors of the current macro block of the current frame is a zero motion vector; (C) determining non-variability between the current macroblock and the corresponding macroblock of the reference frame when the condition of step (b) is satisfied; And And encoding (d) by the direct mode when the current macroblock and the corresponding macroblock are non-variable. The method of claim 5, The direct motion vector of the current macro block of step (b) includes a motion vector with a first reference frame and a direct motion vector with a second reference frame. The method of claim 6, In step (c), the non-variability of the current macroblock and the first corresponding macroblock of the first reference frame and the non-variability of the current macroblock and the second corresponding macroblock of the second reference frame are determined. Fast intra mode determination method for H.264 encoding, characterized in that. The method of claim 7, wherein Whether or not the step (c) is invariant is determined using the Sum of Absolute Difference (SAD) method.

Images