基于VOP的块特性的自适应十字搜索模式运动估计法

Algorithm of adaptive rood pattern search using characteristic of different kinds of block in VOP

提出的利用VOP的块特性的自适应十字搜索模式运动估计法(ARPS-CB-VOP)充分利用VOP的块的特性,不同的块采用不同的运动估计策略。VOP的边框内块分为透明块、边界块和模糊块。当前块为透明块时,不进行运动估计,由解码端直接生成。当前块为边界块时,参考帧不填充,边界块间的SAD计算采用二值α平面值计算。当前块为模糊块时,模糊块的SAD计算采用灰度值计算。另外,对进行运动估计的块,先排除静止块,对运动块,采用自适应十字搜索模式。利用ARPS-CB-VOP法极大地提高了搜索效率,同时计算量显著减少,并且具备较好的峰值信噪比,是一种较有潜力的运动估计和补偿算法。

An algorithm of adaptive rood pattern search using the characteristic of different kinds of block （ARPS-CB- VOP） were proposed, in which the different strategy was used for different kinds of block. The blocks in VOP bounding box were classified into transparent blocks, boundary blocks, and opaque blocks. For transparent blocks, motion estimation was not needed, and was generated by the decoder. For boundary blocks, padding process in the referenced frame was not required, and its SAD （sum of absolute difference） was computed by the binary alpha planes＇ values. For opaque blocks, its SAD was computed by the pixels＇ intensities. Besides, for blocks with motion estimation, static blocks were prejudged, for moving blocks, an adaptive rood pattern search was used. The proposed ARPS-CB-VOP method improves greatly the efficiency of searching, and also the computation load is reduced remarkably with good PRSN （peak ratio of signal to noise）, which is very potential.