摘要
在分析DT(Delaunay triangulation)网格已有的2种分割准则——灰度误差最小化准则和灰度分布均匀化准则在网格基编码不足的基础上,提出了一种新的DT网格分割准则——灰度误差平方和最小化准则,使三角形分割更加合理有效,编码性能比以前的方法有明显提高。在此基础上提出了帧内变换编码/帧间网格基运动估计的混合编码方案,该方案可应用于复杂运动和背景的视频序列中,突破了以往模型基编码只能应用于视频会议的限制。实验结果表明,该编码方案在复杂运动和背景的应用场合高出了H.263的编码性能。
Based on the analysis of the deficiency of two previous Delaunay triangular mesh segmentation criteridifference of gray and minimize variance of gray, a new triangular mesh criterion, minimize sum of squared difference of gray, was proposed. The new segmentation was more rational and efficient so that the coding performance on it was improved obviously than before. Furthermore, a new hybrid coding scheme was also proposed under the criterion with transformation coding for intra frame and mesh-based motion estimation for inter frame, which could be applied in video sequence with complex motion and background. The restriction that mesh model based coding was only applicable in video conference applications was broken. Experimental result shows that the proposed scheme has a better performance than H.263 in applications with complex motion and background.
出处
《通信学报》
EI
CSCD
北大核心
2009年第5期88-93,共6页
Journal on Communications
基金
国家自然科学基金资助项目(60472083,60872141)~~
