KR100896978B1 - The Motion Vector Prediction Method and The Mode Prediction Method and Recording Medium Thereof in Fast MPEG-2 to H.264/AVC Transcoding - Google Patents

Abstract

1. TECHNICAL FIELD OF THE INVENTION

The present invention relates to a motion vector prediction method and a mode prediction method in fast transcoding from MPEG-2 to H.264 / AVC, and to a computer-readable recording medium having recorded thereon a program for realizing each of the above methods.

2. The technical problem to be solved by the invention

According to the present invention, an optimum corresponding to H.264 macroblocks used for fast transcoding based on macroblock type information, coded block pattern information, and motion vector information extracted from the macroblock header information by decoding the MPEG-2 encoded bit stream A computer-readable program recording the motion vector prediction method and the mode prediction method in MPEG-2 to H.264 / AVC, which predicts the motion vector and predicts the optimal mode. The purpose is to provide a record medium.

3. Summary of Solution to Invention

A motion vector prediction method comprising: extracting macroblock type information, encoded block pattern information, and motion vector information from macroblock header information by decoding a bit stream encoded by MPEG-2; Determining a distance between an arbitrary frame and a reference frame based on the extracted macroblock type information, coded block pattern information, and motion vector information; Calculating a motion vector value according to forward prediction, a motion vector value based on backward prediction, a motion vector value based on fast backward prediction, and a motion vector value based on zero prediction using the determined distance information; And determining a motion vector having a minimum motion cost among the calculated motion vector values as an optimal motion vector.

4. Important uses of the invention

The present invention is used for video transcoding and the like.

MPEG-2, H.264 / AVC, Fast Transcoding, Macroblock (MB), Speech Vector Prediction, Mode Prediction

Description

Motion Vector Prediction Method and The Mode Prediction Method and Recording Medium Thereof in Fast MPEG-2 to H.264 / in Fast Transcoding from MP-2 to H.264 / ACC AVC Transcoding}

1 is a configuration diagram of an embodiment of a transcoder to which the algorithm proposed in the present invention is applied.

2 is an exemplary diagram for determining a distance between an arbitrary frame and a reference frame according to the present invention.

3 is a diagram illustrating an embodiment of a backward motion vector (BwdMV), a dominant macroblock (Dominant MB), and a motion vector offset (OffsetMV) according to the present invention.

4 is a flowchart illustrating a motion vector prediction method in fast transcoding from MPEG-2 to H.264 / AVC according to the present invention.

5 is a flowchart illustrating a mode prediction method in fast transcoding from MPEG-2 to H.264 / AVC according to the present invention.

6A-6D are graphs of one embodiment of RD performance of the present invention.

The present invention relates to a fast transcoding method from MPEG-2 to H.264 / AVC and a computer readable recording medium having recorded thereon a program for realizing the method. More specifically, the present invention relates to a bit stream encoded in MPEG-2. Predicts the optimal motion vector corresponding to the H.264 macroblock used for fast transcoding based on the macroblock type information, the coded block pattern information, and the motion vector information extracted from the macroblock header information by decoding The present invention relates to a motion vector prediction method and a mode prediction method in high-speed transcoding from MPEG-2 to H.264 / AVC, and to a computer-readable recording medium having recorded thereon a program for realizing each method.

Video transcoding refers to a technique for converting characteristics of an encoded video stream, for example, bit rate, frame rate, spatial resolution, and encoding standard, to other characteristics. High-speed transcoding in H.264 / AVC is the subject of the technique.

Recently, thanks to the development of the network, users can receive various contents in real time anytime and anywhere. On the other hand, the content providers have different networks for different video contents of the same contents due to different heterogeneous network types. There was a problem in that each of the video content in different formats to produce a protocol suitable for the.

For example, in digital broadcasting systems such as digital satellite broadcasting and digital cable broadcasting, MPEG-2 has already been widely used as a video standard, but recently, H.264 / AVC, which has better video coding rate and image quality, It is gradually adopted in terminals and the like. This not only causes the contents of MPEG-2 contents to be killed, but also generates considerable costs for producing each of the contents conforming to heterogeneous network standards.

As the above problems arise, a technique for transcoding from MPEG-2 to H.264 / AVC has been proposed, and a technique for transcoding at high speed so as to provide video content to a user in real time has been in the spotlight.

Looking at the prior art on transcoding from MPEG-2 to H.264 / AVC, the "Motion and Mode Mapping for MPEG-2 to H." presented in "IEEE Int'l Conference Multimedia & Expo, Toranto Canada, July 2006". 264 / AVC Transcoding "(hereinafter referred to as" prior art 1 ") and" Fast mode decision and motion estimation for H.264 with a "presented in" IEEE Int'l Symp. Circuits Syst., Kobe, Japan, May 2005 ". focus on MPEG-2 / H.264 transcoding "(hereinafter referred to as" prior art 2 ").

In the prior art 1, a transcoding algorithm for converting MPEG-2 broadcast content into an H.264 baseline is proposed. In particular, a motion vector of an MPEG-2 B slice is mapped to a motion vector of an H.264 P slice. Mapping reduces the transcoding complexity.

In the prior art 2, an algorithm for determining a fast mode using neighboring modes information, residue information, and distortion information, and predicting a fast motion vector using a motion vector of MPEG-2 is presented. .

However, such prior arts are difficult to implement optimal performance in terms of tradeoff of rate-distortion. In particular, these prior arts only perform mapping between motion vectors, which has a problem of increasing bit rate and degrading image quality due to inaccurate motion vector use.

The present invention has been proposed to solve the above problems and to meet the above requirements, and includes macroblock type information, coded block pattern information, and motion extracted from macroblock header information by decoding an MPEG-2 encoded bit stream. A motion vector prediction method in fast transcoding from MPEG-2 to H.264 / AVC, which predicts an optimal motion vector corresponding to H.264 macroblocks used for fast transcoding based on vector information and predicts an optimal mode. And a computer-readable recording medium having recorded thereon a mode prediction method and a program for realizing each of the above methods.

Other objects and advantages of the present invention can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

In order to achieve the above object, a method of predicting a motion vector in H.264 / AVC transcoding in MPEG-2 according to the present invention is a motion vector prediction method, which decodes a bit stream encoded in MPEG-2. Extracting macroblock type information, coded block pattern information, and motion vector information from the macroblock header information; Determining a distance between an arbitrary frame and a reference frame based on the extracted macroblock type information, coded block pattern information, and motion vector information; Calculating a motion vector value according to forward prediction, a motion vector value based on backward prediction, a motion vector value based on fast backward prediction, and a motion vector value based on zero prediction using the determined distance information; And determining a motion vector having a minimum motion cost among the calculated motion vector values as an optimal motion vector.

The method may further include determining a final motion vector by performing a refinement step on the determined motion vector.

On the other hand, the method of predicting the mode for high-speed transcoding in MPEG-2 according to the present invention in H.264 / AVC is, in the mode prediction method, I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode. Determining a Mode4 mode, a Mode5 mode, a Mode6 mode, and a Mode7 mode; Decoding the MPEG-2 encoded bit stream to extract macroblock type information, encoded block pattern information, and motion vector information from the macroblock header information; And from among the I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode, Mode4 mode, Mode5 mode, Mode6 mode and Mode7 mode based on the extracted macroblock type information, coded block pattern information and motion vector information. Predicting one or more modes as H.264 macroblock mode.

On the other hand, the present invention provides a transcoder equipped with a processor, the function of decoding the MPEG-2 encoded bit stream to extract macroblock type information, encoded block pattern information and motion vector information from the macroblock header information; Determining a distance between an arbitrary frame and a reference frame based on the extracted macroblock type information, coded block pattern information, and motion vector information; Calculating a motion vector value according to forward prediction, a motion vector value based on backward prediction, a motion vector value based on fast backward prediction, and a motion vector value based on zero prediction using the determined distance information; And a computer readable recording medium having recorded thereon a program for realizing a function of determining a motion vector having a minimum motion cost among the calculated motion vector values as an optimal motion vector.

On the other hand, the present invention, a transcoder equipped with a processor, the function to determine the I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode, Mode4 mode, Mode5 mode, Mode6 mode and Mode7 mode; Extracting macroblock type information, encoded block pattern information, and motion vector information from macroblock header information by decoding a bit stream encoded by MPEG-2; And from among the I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode, Mode4 mode, Mode5 mode, Mode6 mode and Mode7 mode based on the extracted macroblock type information, coded block pattern information and motion vector information. A computer readable recording medium having recorded thereon a program for realizing the function of predicting one or more modes as an H.264 macroblock mode.

The above objects, features and advantages will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, whereby those skilled in the art may easily implement the technical idea of the present invention. There will be. In addition, in describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of an embodiment of a transcoder to which the algorithm proposed in the present invention is applied.

As shown in FIG. 1, a transcoder to which an algorithm proposed in the present invention, such as an algorithm for predicting a motion vector and an algorithm for predicting a mode, is applied in fast transcoding from MPEG-2 to H.264 / AVC, MPEG-2 Decoder 11, Motion Vector and Mode Mapping Unit 12, Post Processing Unit 13, Frame Prediction Inter / An intra prediction unit 14, a DCT quantizer 15, a DCT dequantizer 16, a deblocking filter 17, and an entropy coding unit 18 are included.

In particular, in the present invention, in the fast transcoding of the MPEG-2 encoded bit stream into the H.264 / AVC encoded bit stream in the transcoder, the motion vector / in order to reduce the transcoding complexity and optimize the rate distortion cost. Macro block type information extracted by the mode prediction unit 12 from the MPEG-2 decoder 11 by decoding the MPEG-2 encoded bit stream (MPEG-2 Bitstream) from the macroblock header information (MPEG-2 MB Information), Based on the coding block pattern information and the motion vector information, an optimal motion vector corresponding to the H.264 macroblock is predicted and an optimal mode is predicted.

Meanwhile, in the transcoder, an I frame (hereinafter referred to as an "I frame") of the MPEG-2 bit stream is selected as an eye slice of the H.264 bit stream in accordance with the characteristics of the H.264 baseline. I slices) (hereinafter referred to as "I slices"), and P frames (hereinafter referred to as "P frames") and B frames (hereinafter referred to as "B frames") of the MPEG-2 bit stream. Transcode into P slices (hereinafter referred to as "P slices") of the H.264 bit stream. In addition, a detailed description of known components in the transcoder is considered to obscure the gist of the present invention, and thus description thereof will be omitted. Hereinafter, an algorithm for predicting a motion vector of the present invention will be described in detail. Shall be.

2 is a diagram illustrating an embodiment for determining a distance between an arbitrary frame and a reference frame according to the present invention.

In the algorithm for predicting the motion vector of the present invention, a distance between an arbitrary frame and a reference frame is determined as shown in FIG. 2.

The distance between one frame, for example, the first P frame and an I frame or a next P frame that is a reference frame belonging to the forward direction of the P frame, is defined as InputFwdDistance (hereinafter referred to as "InputFwdDistance"). . Similarly, the distance between a B frame (for example, a B1 frame which is a first B frame or a B2 frame which is a second B frame) and an I frame or a P frame which is a reference frame belonging to the forward direction of each of these B frames is also defined as InputFwdDistance.

The distance between the B frame and the reference frame belonging to the backward direction of the B frame, for example, the distance between the B1 frame and the P frame, and the distance between the B2 frame and the P frame, is inputBwdDistance (hereinafter referred to as “InputBwdDistance”). Is defined.

Referring to FIG. 2, for example, when the GOP (Group of Pictures) structure of the MPEG-2 bit stream is "M = 3", the InputFwdDistance of the P frame is "3", and the InputFwdDistance of the B1 frame, which is the first B frame. Is " 1 " and InputBwdDistance is " 2 ", InputFwdDistance is " 2 " and InputBwdDistance is " 1 "

Next, in the state where the distance between the arbitrary frame and the reference frame is determined as described above, the motion vector value according to the forward predictor of the present invention and the motion vector value according to the backward prediction And a motion vector value according to Fast Backward Predictor and a motion vector value according to known Zero Predictor. Here, known zero prediction is calculated by simply processing the prediction vector as (0, 0).

First, a motion vector value according to the forward predictor of the present invention is calculated as in Equation 1 below.

That is, when the macroblocks of the B frame and the P frame belong to an inter macroblock, the forward motion vector of the macroblock is defined as "FwdMV", and the motion vector value according to the forward prediction is defined. Is defined as "FwdPredMV", and the motion vector value [FwdPredMV] according to forward prediction is calculated by dividing the forward motion vector [FwdMV] by InputFwdDistance.

FwdPredMV = FwdMV / InputFwdDistance

And the motion vector value according to the backward prediction (Backward Predictor) of the present invention is calculated as shown in [Equation 2].

BwdPredMV = {BwdMV + DominPMV-OffsetMV} / InputFwdDistance

In Equation 2, parameters required for calculating a motion vector value according to backward prediction, for example, "DominPMV" and "OffsetMV", will be described first with reference to FIG. 3.

3 is a diagram illustrating an embodiment of a backward motion vector (BwdMV), a dominant macro block (Dominant MB), and a motion vector offset (OffsetMV) according to the present invention.

As shown in FIG. 3, "DominPMV" in [Equation 2] is the most dominant macroblock at the position indicated by the corresponding backward motion vector [BwdMV] (hereinafter, referred to as "dominant MB block"). ) "Means a motion vector. In the present invention, the macroblock of the P frame in which the positions of the macroblocks overlap most is set as a dominant macroblock [Dominant MB]. Here, the motion vector offset OffsetMV represents a difference value between the position of the backward motion vector [BwdMV] and the position of the dominant macroblock [Dominant MB].

Then, returning to Equation 2, if the Inter Macroblock to which the B frame belongs has backward prediction (for example, on a bit stream encoded by MPEG-2), the backward of this macroblock is present. A motion vector is defined as "BwdMV" and a motion vector value according to backward prediction is defined as "BwdPredMV", and the motion vector of the backward motion vector [BwdMV] and the dominant macroblock [Dominant MB] is defined. DominPMV] is added and the value obtained by subtracting the motion vector offset [OffsetMV] is divided by InputFwdDistance to calculate the motion vector value [BwdPredMV] according to the backward prediction.

Then, the motion vector value according to the fast backward prediction (Fast Backward Predictor) of the present invention is calculated as shown in Equation 3 below.

In other words, the motion vector [FastBwdPredMV] according to the fast backward prediction can be calculated using the backward motion vector [BwdMV]. Compute the motion vector [FastBwdPredMV].

FastBwdPredMV =-(BwdMV / InputBwdDistance)

In addition, a motion vector value according to a known zero predictor is also calculated.

Then, the minimum motion cost among the motion vector value according to the forward prediction, the motion vector value according to the backward prediction, the motion vector value according to the fast backward prediction, and the motion vector value according to the zero prediction calculated as described above. The best predictor is determined as a motion vector having the optimal motion vector. Here, the motion cost refers to the cost of the pixel to be encoded, for example, by subtracting the block from the portion to be encoded and the (x, y) position inside the search region when the motion vector is (x, y), and the minimum motion The process of finding the cost is to find the (x, y) with the smallest result by selecting pixels x pixels and subtracting the absolute value, and selecting the motion vector.

In addition, a refinement step may be performed on the motion vectors determined (predicted) as described above, and a known small diamond search pattern (SDSP) commonly used at the position of the optimal motion vector may be used. Patterns can be used to refine and predict this motion vector more accurately. Here, the refining process is to find the optimal position by several times finding the cost of moving (x, y) to the peripheral position, for example, the diamond pattern.

The algorithm for predicting a motion vector in fast transcoding to H.264 / AVC in MPEG-2 according to the present invention described with reference to FIGS. 1 to 3 follows the flow as shown in FIG. 4.

FIG. 4 is a flowchart illustrating a motion vector prediction method in fast transcoding from MPEG-2 to H.264 / AVC according to the present invention.

First, by decoding a MPEG-2 encoded bit stream (MPEG-2 Bitstream) (401), macroblock type information, encoded block pattern information, and motion vector information are extracted from macroblock header information (MPEG-2 MB Information). (402).

Thereafter, a distance between an arbitrary frame and a reference frame is determined based on the extracted macroblock header information (403). For example, the distance between one frame and the reference frame belonging to the forward direction of this frame is defined as "InputFwdDistance", and for the B frame, the distance between the B frame and the reference frame belonging to the backward direction of this B frame is referred to as "InputBwdDistance". define.

Then, the motion vector value according to the forward predictor (Forward Predictor) using [Equation 1] proposed in the present invention, the motion vector value according to the backward prediction (Backward Predictor) using [Equation 2], [ Equation 3] is used to calculate a motion vector value according to Fast Backward Predictor, and calculates a motion vector value according to a known zero predictor (404).

Then, the minimum motion cost among the calculated motion vector values according to the forward prediction, the motion vector values according to the backward prediction, the motion vector values according to the fast backward prediction, and the motion vector values according to the zero prediction are determined. The predicted motion vector is determined (predicted) as an optimal motion vector [best predictor] (405).

In addition, a refinement step may be performed on the motion vectors determined (predicted) as described above, and a known small diamond search pattern (SDSP) commonly used at the position of the optimal motion vector may be used. The determined motion vector may be more accurately refined and determined using a pattern) (406).

Meanwhile, an algorithm for predicting mode in fast transcoding to MPEG-2 according to the present invention will be described in detail.

Macroblock header information can be extracted from the MPEG-2 bit stream decoded by the MPEG-2 decoder 11 of the transcoder shown in FIG. 1, which includes macroblock type information, coded block pattern information, and the like. Motion vector information and the like are included.

Through the macroblock type information, whether the macroblock predicted from the reference frame [Intra or Inter], whether the value of the motion vector is (0, 0) [MC or No-MC], and forward motion It is possible to know whether it is a forward or backward motion prediction or an interpolated (bidirectional) motion prediction. Here, different types of macroblock type information are provided according to frame structures, for example, I frames, P frames, and B frames.

The coded block pattern information may determine whether 8 × 8 blocks in a macroblock are encoded [Coded or NotCoded].

The aforementioned macroblock type information and encoded block pattern information of MPEG-2 provide important information for macroblock mode prediction (selection) of H.264 in high-speed transcoding from MPEG-2 to H.264 / AVC. Based on such macroblock type information and coded block pattern information, the present invention predicts (selects) an optimal mode for high-speed transcoding from MPEG-2 to H.264 / AVC.

Next, an algorithm for predicting a mode in high-speed transcoding in MPEG-2 of the present invention to H.264 / AVC will be described with reference to FIG. 5.

5 is a flowchart illustrating a mode prediction method in fast transcoding from MPEG-2 to H.264 / AVC according to the present invention.

First, the mode used in the present invention is determined as follows (501).

In the intra macroblock mode of H.264, the intra 16x16 prediction mode is set to "I16MB mode", the intra 4x4 prediction mode is set to "I4MB mode", and the inter prediction mode is determined. Here, the inter prediction mode may have various modes according to the characteristics of H.264 using variable block size motion estimation. The selectable block sizes are “SKIP” and “16 × 16”. And "16x8" and "8x8", correspondingly to "SKIP mode", "Mode1 mode", "Mode2 mode" and "Mode3 mode", and "P8x8 mode". In the " P8x8 mode ", the " 8x8 ", " 8x4 ", " 4x8 " and " 4x4 " sizes for each 8x8 sub-macroblock, respectively. Inter prediction is possible, and correspondingly, "Mode4 mode", "Mode5 mode", "Mode6 mode" and "Mode7 mode" are respectively determined.

With each mode determined as described above, the MPEG-2 encoded bit stream (MPEG-2 Bitstream) is decoded (502), and the macroblock type information and the encoding block from the macroblock header information (MPEG-2 MB Information). Pattern information and motion vector information are extracted (503).

Based on the extracted macroblock header information, mode prediction is performed as follows.

That is, the I frame of the transcoding target MPEG-2 frame is predicted (considered) as the H.264 macroblock mode only in the I4MB mode and the I16MB mode except the inter prediction mode.

Meanwhile, for the P frame or the B frame of the transcoding target MPEG-2 frame, when the macroblock of the P frame or the B frame is an intra coded macroblock, I16MB mode, I4MB mode, SKIP mode, and Mode1 mode. Mode 2 and mode 3 are predicted (considered) as an H.264 macroblock mode (505). For example, the reason for performing inter prediction in H.264 despite being an intra macroblock on the MPEG-2 bit stream to be transcoded is fundamentally different from the concept of intra prediction in H.264 and MPEG-2. Since the time required for motion estimation is reduced as much as possible through the motion vector prediction algorithm, the additional time required for the mode prediction is less, and the performance improvement due to this is noticeable.

On the other hand, when the macroblock of the transcoding target MPEG-2 frame is the SKIP mode on the MPEG-2 bit stream, the SKIP mode, the Mode1 mode, the Mode2 mode, and the Mode3 mode are predicted (considered) as the H.264 macroblock mode ( 506). For example, the macroblock of the MPEG-2 frame is the SKIP mode because it means that the block has little information to be encoded.

On the other hand, if the macroblock of the transcoding target MPEG-2 frame is a motion estimation uncoded macroblock (MC), the SKIP mode, Mode1 mode, Mode2 mode, and Mode3 mode are predicted as H.264 macroblock mode. (Consider) (507). For example, a macroblock of an MPEG-2 frame is a motion estimation uncoded macroblock because it means a block having little information to be encoded after inter prediction.

On the other hand, if the macroblock of the transcoding target MPEG-2 frame is a motion non-prediction inter coded macroblock (No-MC), the I4MB mode, the I16MB mode, the SKIP mode, the Mode1 mode, the Mode2 mode, and the Mode3 mode. Predict (consider) H.264 macroblock mode (508). For example, the fact that a macroblock of an MPEG-2 frame is a motion non-prediction inter-encoding macroblock means that the current block has information to be encoded to some extent.

Meanwhile, when the macroblock of the transcoding target MPEG-2 frame is a motion coded macroblock (MC Coded Macroblock), all possible mode prediction is performed (509). For example, it is determined that a macroblock of an MPEG-2 frame is a motion estimation encoded macroblock as a complex region.

In addition, the rate-distortion cost of the I16MB mode is calculated from the H.264 macroblock mode prediction result, and if the rate distortion cost of the I16MB mode is greater than the rate distortion cost of other inter prediction, the H.264 macro It is preferable to exclude the I4MB mode from the block mode prediction result.

As described above, in performing the H.264 mode prediction, it can be seen that most macroblocks have a higher cost of intra prediction than a cost of inter prediction. In addition, in the case of intra prediction, the complexity of the I4MB mode is greater than that of the I16MB mode. With this in mind, the present invention skips the I4MB mode if the rate-distortion cost of the I16MB mode is greater than the rate distortion cost of the other inter predictions, thereby reducing the I4MB mode of intra prediction. Complexity can be reduced.

[Table 1] below summarizes the H.264 mode selected (predicted) for the macroblock of the transcoding target MPEG-2 frame.

Macroblock type of MPEG-2 macroblock candidate in H.264 / AVC Intra Coded MB I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode SKIP MB SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode MC, Not Coded MB SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode No-MC, Inter Coded MB I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode MC, Inter Coded MB All possible modes

6A to 6D are graphs showing one embodiment of RD performance of the present invention, and shields of HD class (1280 × 720 [p], 59.97 [fps], 300 [frames]) which are generally used for comparing RD performance. (shields) images and stockholm images were used.

 FIG. 6A shows the RD performance graphs of the conventional algorithm and the algorithm proposed by the present invention by setting the "search area 32" with respect to the shield image, and FIG. 6B shows the "search area 64" with respect to the shield image. RD performance graphs of the existing algorithms and the algorithms proposed by the present invention are shown, and FIG. 6C shows the search area 32 for the stockholm image. An RD performance graph is shown, and FIG. 6D shows an RD performance graph of each of the existing algorithm and the algorithm proposed in the present invention by setting a "search area 64" with respect to a stockholm image.

In addition, the shields image and the stockholm image were examined by QP from 28 to 40 in units of "4", and compared with experimental results encoded after full decoding while performing the RDO function of H.264. It was.

Table 2 below shows a case where the search area is "32."

Table 3 below shows a case where the search area is "64".

As can be seen from the above [Table 2] and [Table 3] and FIGS. 6A to 6D, the average of the algorithm of the present invention is a search region 32 in terms of time rather than the conventional algorithm that performs encoding after full decoding. You can see that 74.6% is reduced and 88.7% is reduced in the case of 64 search areas. Furthermore, it can be seen that the RD performance of the present invention is hardly different from that of the conventional algorithm that performs encoding after full decoding. It can be seen that the present invention is a very efficient algorithm in terms of time and RD performance compared to the existing algorithm in high-speed transcoding from MPEG-2 to H.264 / AVC.

As described above, the present invention has excellent performance in converting MPEG-2 contents into H.264 / AVC contents, and performs services such as transmitting and storing streaming data such as video and video in real time. The benefit is that it significantly reduces system complexity.

As described above, the method of the present invention may be implemented as a program and stored in a recording medium (CD-ROM, RAM, ROM, floppy disk, hard disk, magneto-optical disk, etc.) in a computer-readable form. Since this process can be easily implemented by those skilled in the art will not be described in more detail.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by the drawings.

The present invention as described above has the effect of significantly reducing the complexity in high speed transcoding from transcoder such as MPEG-2 to H.264 / AVC.

In addition, the present invention has an effect of predicting a motion vector optimized for rate distortion cost by efficiently reusing macroblock header information such as motion vector information of MPEG-2.

In addition, the present invention has the effect that it can take excellent time to transcode from MPEG-2 to H.264 / AVC while achieving superior RD performance.

Claims (18)

In the motion vector prediction method, Decoding the MPEG-2 encoded bit stream to extract macroblock type information, encoded block pattern information, and motion vector information from the macroblock header information; Determining a distance between an arbitrary frame and a reference frame based on the extracted macroblock type information, coded block pattern information, and motion vector information; Calculating a motion vector value according to forward prediction, a motion vector value based on backward prediction, a motion vector value based on fast backward prediction, and a motion vector value based on zero prediction using the determined distance information; And Determining a motion vector having a minimum motion cost among the calculated motion vector values as an optimal motion vector Motion vector prediction method comprising a. The method of claim 1, Determining a final motion vector by performing a refinement step on the determined motion vector; Motion vector prediction method further comprising. The method of claim 2, The refining process performed on the determined motion vector, And a motion vector is refined using a small diamond search pattern (SDSP) at the position of the determined motion vector. The method according to any one of claims 1 to 3, The process of determining the distance between the arbitrary frame and the reference frame, The distance between one frame and the reference frame belonging to the forward direction of the one frame is defined as "InputFwdDistance", and for the B frame, the distance between the B frame and the reference frame belonging to the backward direction of the B frame is referred to as "InputBwdDistance". Motion vector prediction method, characterized in that it is defined. The method of claim 4, wherein The process of calculating the motion vector value ["FwdPredMV"] according to the forward prediction is calculated using Equation 1 below. [Equation 1] FwdPredMV = FwdMV / InputFwdDistance Here, "FwdMV" represents a forward motion vector of the macroblock. The method of claim 4, wherein The process of calculating the motion vector value ["BwdPredMV"] according to the backward prediction is calculated using Equation 2 below. [Equation 2] BwdPredMV = {BwdMV + DominPMV-OffsetMV} / InputFwdDistance Here, "DominPMV" represents a motion vector of a dominant macroblock (hereinafter, referred to as a dominant macroblock [Dominant MB]) at the position indicated by the corresponding backward motion vector ["BwdMV"]. The vector offset ["OffsetMV"] represents the difference between the position of the backward motion vector ["BwdMV"] and the position of the dominant macroblock [Dominant MB]. The method of claim 6, The process of calculating the motion vector value ["FastBwdPredMV"] according to the fast backward prediction is calculated using the following Equation (3). [Equation 3] FastBwdPredMV =-(BwdMV / InputBwdDistance) In a transcoder with a processor, Extracting macroblock type information, encoded block pattern information, and motion vector information from macroblock header information by decoding a bit stream encoded by MPEG-2; Determining a distance between an arbitrary frame and a reference frame based on the extracted macroblock type information, coded block pattern information, and motion vector information; Calculating a motion vector value according to forward prediction, a motion vector value based on backward prediction, a motion vector value based on fast backward prediction, and a motion vector value based on zero prediction using the determined distance information; And A function of determining a motion vector having a minimum motion cost among the calculated motion vector values as an optimal motion vector A computer-readable recording medium having recorded thereon a program for realizing this. The method of claim 8, A function of determining a final motion vector by performing a refinement step on the determined motion vector A computer-readable recording medium that records a program for further realization. In the mode prediction method, Determining an I16MB mode, an I4MB mode, a SKIP mode, a Mode1 mode, a Mode2 mode, a Mode3 mode, a Mode4 mode, a Mode5 mode, a Mode6 mode, and a Mode7 mode; Decoding the MPEG-2 encoded bit stream to extract macroblock type information, encoded block pattern information, and motion vector information from the macroblock header information; And One of I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode, Mode4 mode, Mode5 mode, Mode6 mode and Mode7 mode based on the extracted macroblock type information, coded block pattern information and motion vector information Predicting the above mode as an H.264 macroblock mode Mode prediction method comprising a. The method of claim 10, Predicting the H.264 macroblock mode, A mode prediction method comprising predicting only I4MB mode and I16MB mode as an H.264 macroblock mode except for I frame among MPEG-2 frames to be transcoded. The method of claim 10, Predicting the H.264 macroblock mode, For P or B frames among the MPEG-2 frames to be transcoded, if the macroblock of the P frame or the B frame is an intra coded macroblock, I16MB mode, I4MB mode, SKIP mode, Mode1 mode, and Mode2 mode And predicting Mode3 mode as an H.264 macroblock mode. The method of claim 10, Predicting the H.264 macroblock mode, When the macroblock of the MPEG-2 frame to be transcoded is the SKIP mode on the MPEG-2 bit stream, the mode prediction method comprising predicting the SKIP mode, the Mode1 mode, the Mode2 mode, and the Mode3 mode as the H.264 macroblock mode. . The method of claim 10, Predicting the H.264 macroblock mode, When the macroblock of the transcoding target MPEG-2 frame is a motion estimation uncoded macroblock (MC), the SKIP mode, Mode1 mode, Mode2 mode, and Mode3 mode are predicted as H.264 macroblock mode. Mode prediction method. The method of claim 10, Predicting the H.264 macroblock mode, If the macroblock of the MPEG-2 frame to be transcoded is a motion non-prediction inter coded macroblock (No-MC), the I4MB mode, the I16MB mode, the SKIP mode, the Mode1 mode, the Mode2 mode, and the Mode3 mode are set to H. 264 A mode prediction method characterized in that the prediction as a macroblock mode. The method of claim 10, Predicting the H.264 macroblock mode, If the macroblock of the MPEG-2 frame to be transcoded is an MC coded macroblock, I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode, Mode4 mode, Mode5 mode, Mode6 mode And predicting Mode7 mode as an H.264 macroblock mode. The method according to any one of claims 10 to 16, The rate-distortion cost of the I16MB mode is calculated from the H.264 macroblock mode prediction result, and if the rate distortion cost of the I16MB mode is greater than the rate distortion cost of another inter prediction, the H.264 macroblock mode Mode prediction method, characterized in that the I4MB mode is excluded from the prediction result. In a transcoder with a processor, A function of determining the I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode, Mode4 mode, Mode5 mode, Mode6 mode and Mode7 mode; Extracting macroblock type information, encoded block pattern information, and motion vector information from macroblock header information by decoding a bit stream encoded by MPEG-2; And One of I16MB mode, I4MB mode, SKIP mode, Mode1 mode, Mode2 mode, Mode3 mode, Mode4 mode, Mode5 mode, Mode6 mode and Mode7 mode based on the extracted macroblock type information, coded block pattern information and motion vector information Function to predict the above modes as H.264 macroblock mode A computer-readable recording medium having recorded thereon a program for realizing this.

Images