摘要
提出了一种新的新闻视频静态图形标识分割算法.采用时域采样选取代表帧序列,利用静态标识的时空特性定位潜在标识区域,结合标识区域的尺寸、位置等特征滤除干扰区域,最终分割出标识区域.与已有方法相比,所提方法实现简单,可以有效分割半透明标识.
A novel static graphic logo localization algorithm for news videos was developed. A temporal sample based technique is employed to select representative frames, The spatial-temporal characteristics of static logo are used to locate potential logo areas. Finally, several restrictions, including the size and location of each candidate logo area, are applied to eliminate the non-logo regions. Compared with the previous methods, the proposed approach is computationally efficient, and can effectively segment semi-transparent logos.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2006年第5期758-761,共4页
Journal of Shanghai Jiaotong University
