| MPEG-4 is a standard of object-based multimedia data compression and coding, and belongs to the second era content-based moving picture standard. The application of MPEG-4 is very wide. MPEG-4 is a very open standard, which allows competition and improvement for the open parts, such as video segmentation, rate control and coding technology and so on. The key technologies of MPEG-4 include the image analysis and segmentation technology and the coding technology, which are becoming the hotspot of study and development.This paper studies on the key technologies of MPEG-4, in which the task accomplished includes two parts, as follow:The first part is the one of video segmentation approaches and algorithms for MPEG-4. The representative methods and algorithms of VOP (Video Object Plane) generation are discussed, and the idea of method, the content and process of algorithm and the advantage and disadvantage of algorithm are analyzed. The approach of automatic segmentation of VOP's based in spatio-temporal information (SBSTI) is proposed, and is discussed in details. Segmentation Based in Temporal Information (ST) can segment the foreground with a fast or slow motion by the feature of motion and multiple frames, and also overcome the shortcoming of error segmentation because of occlusion, random noise, coarse motion estimation of regions at the boundary of the object. Segmentation Based in Spatial Information (SS) can segment an image into different regions by the feature of hue in the single frame. Segment Fused Spatial and Temporal Information (SFST) can take the advantage of ST to provide coarse mask of moving VOP and the advantage of SS to provide accurate boundary, and also overcome the shortcoming of ST to provide too coarse boundary resulting in coarse segmentation and shortcoming of SS to segment image into many regions resulting in over-segmentation. The model of VOP is initiated simply and efficiently. The updated model can track rotated, distorted and still VOP. The detection of shot break can achieve segmentation of whole video sequence.The second part is the one of video coding methods and algorithms for MPEG-4.The paper studies the coding of VOP's shape, texture, and motion information respectively, which includes as follow:The part of the approaches and algorithms of VOP's shape coding for MPEG-4.The representative methods and algorithms of VOP's shape coding are discussed, and the idea of method, the content and process of algorithm and the advantage and disadvantage of algorithm are analyzed. The modified quad-tree mutli-resolution shape coding algorithm is proposed, in which the complex of quad-tree is controlled by the homogenous parameter, so as to improve the efficiency of shape coding. The motion is estimated by searching only in the efficient search areas during the inter-frame shape coding of VOP, so as to improve greatly the efficiency of searching.The part of the approaches and algorithms of VOP's texture coding for MPEG-4.The representative methods and algorithms of VOP's texture coding are discussed, and the idea of method, the content and process of algorithm and the advantage and disadvantage of algorithm are analyzed.The part of the methods and algorithms of VOP's motion estimation and compensation for MPEG-4.The representative approaches and algorithms of VOP's motion estimation and compensation for MPEG-4 are discussed, and the idea of method, the content and process of algorithm and the advantage and disadvantage of algorithm are analyzed. The algorithm of adaptive rood pattern search using the characteristic of different kinds of block is proposed, in which the different strategy is used for different kinds of block. First, the blocks of VOP bounding box is classified into the transparent blocks, the boundary blocks, and the opaque blocks. For the transparent blocks, the motion estimation is not needed, and is generated by the decoder. For the boundary blocks, the padding process in the referenced frame is not required, and its SAD (the sum of absolute dif...
|
PDF Full Text Request