KR0178202B1 - Method for establishing the window in motion estimation - Google Patents

Abstract

The present invention relates to a method for setting a window in motion estimation. The present invention relates to calculating a macroblock position of a B frame existing on a line connected between I frame and P frame image data, and configuring a search area A for the calculated position. By setting the window size on the I-frame and P-frame image data according to the prediction direction, and then performing the motion estimation, the window size is properly adjusted during the search of the macroblock for motion estimation, thereby reducing the computation time and accurate The motion vector can be calculated.

Description

Window setting method in motion estimation

1 is a schematic block diagram of a typical typical digital video encoder.

2 is a flowchart illustrating an operation of adjusting the size of a window according to a preferred embodiment of the present invention.

3 is a view for explaining the setting of the size of the window according to the present invention.

* Explanation of symbols for main parts of the drawings

110: subtractor 120: DCT unit

130: quantization unit 140: inverse quantization unit

150: IDCT unit 160: adder

170: frame memory 180: motion estimation unit

190: motion compensation unit

The present invention relates to a motion estimation method, and more particularly, to a method for setting a window in motion estimation, in which a motion estimation is performed by adjusting the size of a window during motion estimation.

In general, when video data is to be transmitted as a digital signal such as a video telephone or HDTV, digital data transmitted by using a high rate data compression method is compressed, that is, encoded and transmitted in order to reduce the enormous amount of data. do.

In encoding such video data, the most important thing is to reduce the amount of data to be transmitted. Therefore, the transmitting side encodes the moving image data using the spatial and temporal correlation of the image data and then transmits the moving image data to the receiving side through the transmission channel.

More specifically, the encoding system on the transmitting side removes spatial redundancy of image data by using transform coding such as DISCRETE COSINE TRANSFORM (hereinafter abbreviated as DCT). By eliminating temporal redundancy of the image data using encoding, the image data is efficiently compressed.

On the other hand, frame image data captured by the camera and converted into digital image data are provided in the order of I, B, B, P, B, B, P, B, B, P, B, B, I .... This is a frame order of image data in the MPEG (MOTION PICTURE EXPERTS GROUP) standard. A series of frame image data is generated from an I frame coded without motion compensation and a previous I frame or P frame. Or a P frame containing intra frame information, and a B frame generated by motion compensation using I and P frames before and after.

Here, motion estimation searches for a block to be identified using a macroblock between image data of a current frame and image data of an adjacent frame, and performs motion compensation using a displacement of the searched block, that is, a motion vector. .

Meanwhile, in the method of searching for macro blocks that are identified by using a macro block between the image data of the current frame and the image data of an adjacent frame, the frame image data is classified as a window and then divided into the divided windows. We searched the identified blocks of macroblocks to be searched in. However, in the prior art, since the size of the window is predetermined, a large range of macroblocks can constitute an accurate motion vector, but it takes a long operation time and has a complicated hardware implementation. In addition, if the range of the macroblock is small, the operation time is short, but there is a problem that an accurate motion vector cannot be constructed.

Accordingly, the present invention has been made in view of the above-described problems of the prior art, and an object of the present invention is to provide a window adjusting method in motion estimation that can minimize the size of a window when estimating the motion of interpolated frame image data. have.

In order to achieve the above object, the present invention provides a method for estimating a motion vector of I frame and P frame image data through motion estimation means, wherein a window for a huge search region is set and then the I through motion estimation means. A first step of estimating a motion vector with respect to the frame and the P frame; determining a relative position of the macroblock with respect to the B frame image data; and if the prediction direction with respect to the relative position of the macroblock is forward prediction, the I frame image A second step of setting a smaller window in the data than the window for the giant search area; and if the predictive direction with respect to the relative position of the macroblock is backward prediction, the window for the giant search area for the P-frame image data. A third step of setting a relatively small window and the huge search area set in the second and third steps; And a fourth step of estimating the motion vector through the motion estimating means using a window smaller than the window for.

The above and other objects and various advantages of the present invention will become more apparent from the preferred embodiments of the present invention described below with reference to the accompanying drawings by those skilled in the art.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a schematic block diagram of a typical digital video encoder. The subtractor 110, the DCT unit 120, the quantizer 130, the inverse quantizer 140, the IDCT unit 150, and the adder 160 are shown in FIG. ), A frame memory 170, a motion estimator 180, and a motion compensator 190.

In FIG. 1, the subtractor 110 subtracts the input frame image data and the frame image data compensated by the motion compensator 190 and outputs them to the DCT unit 120, and the DCT unit 120 receives the input time. The frame image data of the region is converted into a transform coefficient of the frequency domain, and the quantization unit 130 quantizes the frame image data as the transformation coefficient input from the DCT unit 120 by approximating a finite number of values.

In addition, the inverse quantization unit 140 and the inverse discrete cosine transform unit (hereinafter, referred to as IDCT unit) 140 convert the frame image data transformed and quantized by the DCT unit 120 and the quantization unit 130. The frame image data is restored to the original frame image data for motion compensation, and the frame memory 170 restores the frame image data restored from the IDCT unit 150 and the motion compensated frame image data added through the adder 160 to the previous frame. Save it.

In addition, the motion estimator 180 estimates the motion by using the current frame image data and the previous frame image data stored in the frame memory 170 among the frame image data before the adder 160 to compensate for the motion. The compensator 190 compensates for the motion of the frame image data by moving the image data of the previous frame by the motion vector according to the motion estimation from the motion estimator 180.

In estimating the motion of the previous frame image data and the current frame image data stored in the frame memory 170, the motion estimation unit 180 among the components of the digital image encoder as described above, The operation process of setting will be described in more detail with reference to FIGS. 1 and 2.

First, the motion estimation unit 180 sets a window for the search area B between an I frame and a P frame, obtains a motion vector, and obtains positions of macroblocks of the B frames (steps 210 and 220).

Next, as shown in FIG. 3, the B frame image data calculates the macroblock position of the B frame on the line connected between the I frame and the P frame. Referring to FIG. When the coordinate of the motion vector obtained between the frame and P is X, the position of the macroblock of the frame image data of B1 is (X1, Y1) = (2 / 3X, 2 / 3Y), and the macro of the frame image data of B2. The position of the block is (X2, Y2) = (1 / 3X, 1 / 3Y).

Therefore, by predicting the position of the macroblock obtained in steps 210 and 220 (step 240), in the case of forward prediction, a window for the search area A is set in the I frame image data to construct a motion vector (step 250). In the case of backward prediction, a motion vector is configured by setting a window for the search area A in the P frame image data (step 260).

The motion vector is estimated in the window for the search area A configured in steps 250 and 260 (step 270).

As described above, the position of the macroblock of the B frame existing on the line connected between the I frame and the P frame image data is calculated, and a search area A is formed for the calculated position. After setting the window size on the P frame image data, motion estimation is performed.

Therefore, when the present invention is used, the size of the window is appropriately adjusted during the search of the macroblock for motion estimation, thereby reducing the computation time and calculating the accurate motion vector.

Claims (1)

In the method for estimating the motion vector of the I frame and the P frame image data through the motion estimation means 180, the window for the huge search area is set and then the I frame and the P through the motion estimation means 180. Estimating a motion vector for the frame; Determining a relative position of the macroblock with respect to the B-frame image data, and setting a small window in the I-frame image data as compared to the window for the huge search region when the prediction direction of the macroblock relative to the macroblock is forward predicted. Step 2; Setting a window smaller than the window of the huge search region for the P-frame image data if the prediction direction of the macro block is backward prediction; In the motion estimation, characterized in that the fourth step of estimating the motion vector through the motion estimation means 180 using a window smaller than the window for the giant search area set in the second and third steps How to set Windows.