基于小波域分水岭变换的视频对象提取算法

An Automatic Video Object Extraction Algorithm Based on Watershed in the Wavelet Domain

提出了一种基于时、空分割融合的视频对象分割算法.时间分割采用变化检测, 其关键的阈值选取通过直方图分析得到.空间分割采用小波域的分水岭变换算法.时、空分割融合采用基于区域的双向投影策略.实验结果表明,本算法可比COST211 AM参考分割取得更好的分割结果.

An automatic video object segmentation algorithm based on spatio-temporal informtion was proposed. Temporal segmentation was based on change detection, whose key factor was the selection of appropriate threshold by histogram analysis. For spatial segmentation, watershed transformation was performed in the wavelet domain. The fusion of the spatio-temporal segmentation was based on a bi-directional projection approach. Experimental results have shown that it can achieve more desirable results than COST211 AM.