KR20080107668A - Method for motion estimation based on image block and apparatus thereof - Google Patents

Abstract

A motion estimation method and an apparatus thereof are provided to select a search center value and to improve the flexibility of the motion estimation algorithm due to changing various search algorithms. One or more search central values for the motion estimation toward the predetermined block of the current frame are inputted(310). A reference block corresponding to a search area within a reference frame based on the inputted search central value is searched(340).

Description

Method for motion estimation based on block and apparatus thereof

1 is a diagram illustrating a general video encoding apparatus.

2 is a flowchart illustrating a block-based image motion estimation method according to an embodiment of the present invention.

3 is a functional block diagram illustrating a block-based image motion estimation apparatus according to another embodiment of the present invention.

4 is a diagram illustrating adjacent blocks usable as search medians.

5A to 5C are diagrams illustrating respective cases determined as search areas in motion estimation according to the present invention.

Corresponding reference numerals in the several drawings indicate corresponding parts. Although the drawings show embodiments of the invention, the drawings are not to scale and certain features may be exaggerated to better illustrate and explain the invention.

The present invention relates to a block-based video motion estimation method. More specifically, the motion estimator provides a predefined and changeable structure for the motion estimator that takes up most of the operations in a video encoder. The present invention relates to a motion estimation method and apparatus for enabling various changes.

Motion estimation, which is one of encoding techniques of video data, uses a predetermined measurement function to search the macro frame, which is most similar to the macro block in the current frame, by searching a reference frame. It is a process of obtaining a motion vector representing the difference in the positions of.

The criterion for determining the most similar macroblock is a predetermined measurement function called a matching criterion, which is the step that requires the most computational amount to account for 50% to 90% of the total computation in the encoding process of video data. .

A search process is a process of finding a block most similar to the current frame existing in the reference frame with respect to a search area having a predetermined size and a block having a predetermined size. The search process is divided into full search and local search. The global search is a process of searching for a block having a minimum error value from the current block by measuring energy at all positions in the search area by using a function, and the local search is performed only for a part of the positions in the search area about a predetermined reference point. It is a search process made by measuring energy.

Matching reference values are calculated using a predetermined measurement function, such as Sum of Absolute Differences (SAD), Sum of Absolute Transformed Differences (SATD), Sum of Squared Differences (SSD), Mean Squared Differences (MSD), Mean Absolute Differences (MAD).

The encoding process of such a general image will be described first, and then the problems of the prior art will be described.

1 is a diagram illustrating a general video encoding apparatus.

Referring to FIG. 1, first, the transform unit 110 performs a discrete cosine transform or a Karhuhen-Loeve transform on a video data input in units of blocks in order to remove spatial correlation. Value Decomposition (SVD) operation is performed, and the quantization unit (Quantization) 120 performs quantization on the transform coefficient obtained by the transform unit 110 and expresses it as some representative value, thereby performing high efficiency lossy compression. The variable length coding unit 170 removes statistical redundancy from the DCT and the quantized data.

The inverse quantization unit 130 inverse quantizes the quantized image data in the quantization unit 120, and the inverse transform unit 140 inverses Discrete Cosine for the inverse quantized image data in the inverse quantization unit 130. Perform an inverse transformation, such as Transform). The frame memory unit 150 stores the image data inversely transformed by the inverse transform unit 140 in units of frames.

The motion estimation and compensation unit 160 uses the image data of the current frame to be input and the image data of the previous frame stored in the frame memory unit 150 to obtain a motion vector (MV) per macro block (MB). And SAD (Sum of Absolute Differences) corresponding to a block matching error.

The calculation of the block matching error is very large and requires a lot of hardware resources to implement real-time video data encoding. In order to reduce the amount of computation, many motion estimation techniques have been studied to reduce the amount of computation to be performed in motion estimation. Motion estimation methods include a wide variety of algorithms such as full search algorithm, multi-resolution multiple candidates search (MRMCS), three-step search, and diamond search.

In the calculation of such motion estimation, a motion estimation for processing CIF (Common Intermediate Format) image or SD (Standard Definition) image due to the recent increase in the computing power of the processor such as digital signal processing (DSP) Negative software solutions also appear, but hardware solutions are still chosen for high definition (HD) or high definition coding.

However, hardware solutions have the advantage of real-time processing, but they are less flexible than software, so algorithms such as modifying or updating algorithms after they are implemented as SOC (System On Chip) are implemented. There is a drawback that it is impossible to transform. Therefore, in the case of a hardware solution, it is inconvenient to select and implement one algorithm or to prepare various algorithms to be implemented from the beginning.

Accordingly, the present invention has been devised to solve the above problems, and the technical problem to be achieved by the present invention is to devise a motion estimation method and apparatus having a search center architecture that can be changed by user definition. Even after the motion estimator is implemented as a hardware solution, a method and apparatus for applying various motion estimation algorithms are provided.

According to an aspect of the present invention, there is provided a block-based video motion estimation method, comprising: inputting at least one search center value for motion estimation of a predetermined block of a current frame; And searching for a reference block corresponding to the block of the current frame in the search region in the reference frame based on the input search median.

Inputting a motion vector value of at least one or more pre-coded blocks adjacent to the block of the current frame or a motion vector value of the block at a resolution lower than the original resolution of the image, wherein the searching comprises: Preferably, the reference block is searched based on a search median value and / or the input motion vector value.

According to another aspect of the present invention, in the block-based image motion estimation apparatus, a search region for determining a search region for motion estimation by receiving at least one or more search median values for a predetermined block of a current frame A processing unit; And a block search unit for searching for a reference block corresponding to the block in the search area in the reference frame based on the search median inputted to the search area processor.

A motion vector value of at least one pre-coded block adjacent to the block of the current frame or a motion vector value of the block at a lower resolution than the original resolution of the image is calculated, and the motion vector value is transmitted to the search region processor. It is preferable to further include a search region induction unit.

The search region processor may include: a search region register configured to store the input search median value and / or the received motion vector value; More preferably, the apparatus further includes a memory reading unit which reads data of the current frame and the reference frame from an external memory.

DETAILED DESCRIPTION In order to fully understand the present invention, the operational advantages of the present invention, and the objects achieved by the practice of the present invention, reference should be made to the accompanying drawings which illustrate preferred embodiments of the present invention and the contents described in the drawings.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the present invention.

2 is a flowchart illustrating a block-based image motion estimation method according to an embodiment of the present invention.

Referring to FIG. 2, in the motion estimation method according to the present invention, inputting at least one or more search center values for motion estimation for a predetermined block of a current frame is performed in step 210 and inputting a search center value. A step 230 of searching for a reference block corresponding to the block of the current frame in the search region in the reference frame is performed, and optionally, the motion vector value or at least one pre-coded block adjacent to the block of the current frame. And inputting a motion vector value of the block on a resolution lower than the original resolution of.

In order to understand the operation process of the motion estimation method according to the present invention, the motion estimation apparatus will be described in detail with reference to FIG. 3.

3 is a functional block diagram illustrating a block-based image motion estimation apparatus according to another embodiment of the present invention.

According to the present invention, a motion estimation apparatus generally receives a search area processor 310 and a search area processor 310 for determining a search area for motion estimation by receiving one or more search median values for a specific block of a current frame. A block search unit 340 for searching a reference block corresponding to the current block in the search region in the reference frame based on a median value, and further including a motion vector value or image of one or more pre-coded blocks adjacent to the block of the current frame. The apparatus may further include a search region inducing unit 320 that calculates a motion vector value of the current block at a resolution lower than the original resolution and transmits the motion vector value to the search region processor 310.

The search region processor 310 may include a search region register 311 storing an input search median value and the received motion vector value, and a memory reading unit 312 reading data of a current frame and a reference frame from an external memory. ) May be further included.

Referring to FIG. 3, a function performed for each block will be described. The search region processor 310 calculates a search region for motion estimation. That is, based on a search center value received from an external source such as a user, the reference block for the block to be encoded in the current frame is referred to by referring to the motion vector value of the adjacent block received from the search region inducing unit 320. Determine the search area for searching in the previous frame. The search region processor 310 may include a search region register 311 which stores the search median value and a motion vector value of an adjacent block, and also reads macro block data in a current frame and a reference frame from an external memory. The apparatus may further include a memory reading unit 312 provided to the internal memory 330.

The search region induction unit 320 receives data such as a motion vector value of an adjacent block from an internal memory, calculates a motion vector predictor (MVP) using the same, and also hierarchical motion estimation (MRMCS). A 4: 1 search engine associated with a resolution multiple candidates search may be included to use this to derive a search median. That is, the search region induction unit 320 uses not only the search median directly input from the outside, but also the correlation with neighboring adjacent blocks in the frame, and induces the search median used in the existing fast motion search algorithm. ) Helps to transform the search area in more ways.

The block search unit 340 searches for the search area determined by the search area processor 310 in the horizontal / vertical direction, and the controller 350 generates a control signal for controlling each part of the motion estimation unit.

Looking at the operation of the motion estimation of the present invention based on each of these blocks as follows.

(1) The search region deriving unit 320 stores motion vectors previously obtained in macroblocks adjacent to the current macroblock, and obtains a predicted motion vector (MVP) using the stored motion vectors.

(2) A search center value defined by a user is received from the outside.

(3) The search center values are taken from the obtained search center value candidates in the number and order defined by the user, and stored in the search area register 311 in the search area processor 310.

(4) The data of the search region of the corresponding reference frame is read from the internal memory 330 in the order stored in the search region register 311 and supplied to the block search unit 340.

(5) The block search unit 340 performs a search in the horizontal / vertical range (-w, + w) around the search median value and outputs a motion vector. The obtained motion vector is stored as a motion vector value of an adjacent block in the search region inducing unit 320 for the next operation.

4 is a diagram illustrating adjacent blocks usable as search medians.

Referring to FIG. 4, as is generally known, macroblocks on the left side, upper side, and upper right side of the current macroblock Curr. MB are adjacent blocks and have the greatest correlation with the movement of the current macroblock. In addition, the motion vectors of the macroblocks on the left side (NMV _a ^t ), the upper side (NMV _b ^t ), and the upper right side (NMV _c ^t ) may be used to calculate the prediction value (MVP) of the motion vector of the current macroblock. In addition, the motion vector value of the upper left (NMV _d ^t ) macroblock of the current macroblock may be referred to.

5A to 5C are diagrams illustrating respective cases determined as search areas in motion estimation according to the present invention.

Referring to FIG. 5A, as an embodiment when the search region is [-w, + w] = [-7, +7], four search medians are selected as shown in the figure, respectively (-8, -8). When defined as, (+7, -8), (-8, +7), (+7, +7), the search area of horizontal / vertical [-15, +14] is centered around (0, 0). Branches work the same as full search. That is, four areas searched from four search medians include one large area [-15, +14] centered on (0, 0). Therefore, by applying such a method, it is possible to perform a global search algorithm within a search area having various sizes centering on (0, 0).

Referring to FIG. 5B, unlike FIG. 5A, the search median is (x-8, y-8), (x + 7, y-8), (x-8, y + 7), (x + 7, 4 cases of y + 7). Where (x, y) is the coordinate of the predicted motion vector (MVP). In this case, it operates in the same manner as a full search having a horizontal / vertical [-15, +14] search area centered on MVP, reflecting the predicted movement.

Referring to FIG. 5C, MV ^t _{4 : 1 ( first )} , MV ^t _{4 : 1 ( second )} and a predicted motion vector are shown as search medians. Where MV ^t _{4 : 1 ( first )} and MV ^t _{4 : 1 ( second )} are the motion vectors associated with the hierarchical motion estimation algorithm. The hierarchical motion estimation technique divides the original frame into frames of various resolutions and A technique for generating a motion vector hierarchically with respect to a frame. Among them, a multi-resolution multiple candidate search (MRMCS) is currently known.

That is, MV ^t _{4 : 1 ( first )} and MV ^t _{4 : 1 ( second )} are the best motion vectors obtained by performing subsampling of the original resolution images at 4: 1, respectively, and the second This represents a good motion vector. In this case, since the search median of FIG. 5C includes all search points of the MRMCS algorithm, an algorithm that guarantees performance of the MRMCS algorithm is performed.

Therefore, the motion estimation method according to the present invention may be performed by being transformed into a shape that operates like the various algorithms described above.

Meanwhile, the above-described block-based image motion estimation method of the present invention can be written as a program that can be executed in a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium.

In addition, as described above, the structure of the data used in the present invention can be recorded on the computer-readable recording medium through various means.

The computer-readable recording medium may be a magnetic storage medium (for example, a ROM, a floppy disk, a hard disk, etc.), an optical reading medium (for example, a CD-ROM, a DVD, etc.) and a carrier wave (for example, the Internet). Storage medium).

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

As described above, according to the block-based image motion estimation method and apparatus according to the present invention, although it is a hardware solution for the motion estimation unit, a search center value can be selected differently, and various search algorithms are used. Modifications can be made to improve the flexibility of the motion estimation algorithm.

Claims (5)

In the block-based image motion estimation method, Inputting at least one search center value for motion estimation for a given block of a current frame; And searching for a reference block corresponding to the block of the current frame in a search area within a reference frame based on the input search median value. The method of claim 1, Inputting a motion vector value of at least one or more pre-coded blocks adjacent to the block of the current frame or a motion vector value of the block at a resolution lower than the original resolution of the image; The searching may include searching for the reference block based on the search median value and / or the input motion vector value. In the block-based video motion estimation apparatus, A search region processor configured to receive at least one search median value for a predetermined block of the current frame to determine a search region for motion estimation; And a block search unit for searching for a reference block corresponding to the block in a search area in a reference frame based on a search median inputted to the search area processor. The method of claim 3, A motion vector value of at least one pre-coded block adjacent to the block of the current frame or a motion vector value of the block at a lower resolution than the original resolution of the image is calculated, and the motion vector value is transmitted to the search region processor. Motion estimating apparatus further comprises a search region inducing unit. The method of claim 4, wherein The search region processing unit, A search region register for storing the input search median value and / or the received motion vector value; And a memory reading unit which reads data of the current frame and reference frame from an external memory.

Images