Independent motion is a strong cue that two points belong to different "rigid" entities. This paper presents a method to extract a part-based model of an observed scene from a video sequence. ![]() Index Terms-Block-based affine motion compensation, fast block-matching algorithm, H.264/AVC, high efficiency video coding (HEVC), zoom motion estimation. Experimental results show that an average bit-rate savings of 11.01% (up to 16.87%) can be achieved for high-definition size sequences with an average 0.20 dB gain in peak signal-to-noise ratio and a less than 10% time increment in the hierarchical B structure. We have integrated the proposed algorithm into H.264/AVC JM reference software. For fast ZME, we have also designed a 3-D diamond pattern search to reduce the number of unnecessary search points for the ZV. The resized reference block is then interpolated to the size of the current block in order to perform block matching. To represent and encode the zoom motion, we introduce a zoom vector (ZV) that either increases or decreases a reference block size. In this paper, we present a novel zoom motion estimation (ZME) and motion compensation algorithm using the local area scaling technique. In spite of good performance for these algorithms, they are not used in the current industrial world because of their high computational complexity and incompatibility with current block-based video coding standards that limit their practical usage. ![]() Various algorithms have been proposed for estimating other motion models, such as zooming, rotating, tilting, and warping. However, this traditional ME technique suffers the limitation that it can only compensate for a pure parallel translation between frames. Block-based motion estimation (ME) has been widely used in various video coding standards due to its effectiveness in removing temporal redundancy.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |