FAST MULTI-BLOCK SELECTION FOR H.264 VIDEO CODING Chang, A.; Wong, P.H.W.; Yeung, Y.M.; Au, O.C.; Circuits and Systems, ISCAS '04. Proceedings of the 2004 International Symposium on, Volume: 3, May 2004

Online Introduction of H.264. Introduction of H.264. Introduction of FMBME. Introduction of FMBME. FMBME: a fast multi-block motion estimation. FMBME: a fast multi-block motion estimation. The best match function of multi-block motion estimation. The best match function of multi-block motion estimation. The useful multi-block motion estimation. The useful multi-block motion estimation. Object movement. Object movement. Texture regions. Texture regions. The proposed fast multi-block selection algorithm. The proposed fast multi-block selection algorithm. Experimental result. Experimental result. Conclusion. Conclusion.

Introduction of H.264 The JVT/H.264 is the new video coding standard which additional features concludes: The JVT/H.264 is the new video coding standard which additional features concludes: Intra prediction mode for I-frames. Intra prediction mode for I-frames. Different mode decision including multiple block size. Different mode decision including multiple block size. Macroblock Macroblock Types Types Sub macroblock Sub macroblock Types Types Multiple reference frame motion estimation. Multiple reference frame motion estimation. 16 * 16 4 * 44 * 88 * 48 * 8 8 * 1616 *

Introduction of FMBME We propose a fast multi-block selection scheme : We propose a fast multi-block selection scheme : 16*16 (mode 1), 16*8 (mode 2), 8*16 (mode 3), 8*8 (mode 4) which can efficiently reduce the computational cost while achieving similar visual quality and bit-rate. 16*16 (mode 1), 16*8 (mode 2), 8*16 (mode 3), 8*8 (mode 4) which can efficiently reduce the computational cost while achieving similar visual quality and bit-rate. The sequence of a fast multi-block selection scheme is : The sequence of a fast multi-block selection scheme is : Mode 4 ( 8 * 8 ) Mode 1 ( 16 * 16 ) Mode 2 ( 16 * 8 ) Mode 3 ( 8 * 16 )

The best match function of multi- block motion estimation The best match is found by minimizing the cost function: The best match is found by minimizing the cost function: J(m,λ MOTION )=SAD(s,c(m))+λ MOTION *R(m-p) J(m,λ MOTION )=SAD(s,c(m))+λ MOTION *R(m-p) m=(m x,m y ) T is the motion vector. m=(m x,m y ) T is the motion vector. p=(p x,p y ) T is the prediction for the motion vector. p=(p x,p y ) T is the prediction for the motion vector. λ MOTION is the Lagrangle multiplier. λ MOTION is the Lagrangle multiplier. R(m-p) is the bits rate. R(m-p) is the bits rate. SAD ( Sum of Absolute Differences ): SAD ( Sum of Absolute Differences ): B=16, 8 or 4 is the original video signal. B=16, 8 or 4 is the original video signal. c is the coded video signal. c is the coded video signal.

The useful multi-block motion estimation. There are two situations : Object movement and texture regions. There are two situations : Object movement and texture regions. Object movement : Object movement : This means more than one object in a macroblock and are moving in different directions. This included objects moving over a background with different velocity. Fig. 2 the object is moving against a static background. In this case, the current block should be divided into two 8x16 sub-blocks such that the cost function can be minimized.

The useful multi-block motion estimation. Texture regions: Texture regions: If the edges have a integer-pixel offset as shown in Fig. 3a. If the edges of texture have a half-pixel or a quarter-pixel offset relative to the senor, the edges may be blurred as shown in Fig. 3b, 3c. The pixel at the blurred edges may have only ½, 3/4 or ¼ the intensity of the original one, which can lead to difficulty in motion estimation. Since it is possible for a macroblock to contain more than one kind of texture, using only one integer, half or quarter pixel motion vector will not be sufficient to describe the texture content.

The proposed fast multi-block selection algorithm. The proposed FMBME it analysis the information obtained from 8*8 block size ME to predict the mode 1 macroblock. The proposed FMBME it analysis the information obtained from 8*8 block size ME to predict the mode 1 macroblock. As a result, computation can be saved for mode 2 and mode 3 block size ME and in some situation mode as well. As a result, computation can be saved for mode 2 and mode 3 block size ME and in some situation mode as well. Three decisions are set up in different video area: Three decisions are set up in different video area: general area, slow moving area and fast moving area. general area, slow moving area and fast moving area.

The proposed fast multi-block selection algorithm. Defining MV0, MV1, MV2, MV3 is the motion vector of 8*8 subblock of the macroblock. Defining MV0, MV1, MV2, MV3 is the motion vector of 8*8 subblock of the macroblock. C1: If MV0=MV1=MV2=MV3 then - choose mode 1 (16x16) as final block type - no ME will be further performed - 16x16 MV = 8x8 MV0 C2: If three subblock MV are the same AND the forth unequal MV only differ by one quarter pixel (1/4) distance then - choose mode 1 (16x16) as final block type - no ME will be further performed - 16x16 MV = dominated 8x8 MV

The proposed fast multi-block selection algorithm.

C3: If collocate MB in previous frame is mode1 AND {MV0, MV1, MV2, MV3} < 4 (i.e. one integer pixel distance) AND MV0, MV1, MV2, MV3 has the same direction then - choose mode 1 (16x16) as final block type - 8 point local search around MV = {0, 0}

The proposed fast multi-block selection algorithm.

C4: If all magnitude of 8x8 MVx >= 3 integer distance OR all agnitude of 8x8 MVy >= 3 integer distance - choose mode 1 (16x16) as final block type - local search for surrounding 24 points of MV0

The proposed fast multi-block selection algorithm.

Experimental result.

Conclusion. In this paper, a novel fast multi-block motion estimation called FMBME is proposed for H.264 video coding. Experiment results suggest that FMBME can reduce the computation cost significantly with little change to PSNR and bit rate.