基于HVS和时域感知失真的感兴趣区域搜索

Interesting region search based on HVS and temporal perceptual distortion

视频质量评价模型对于感知视频编码有着重要意义.观察者对于视频场景中不同区域有不同的视觉兴趣性,而度量不同区域人眼感兴趣程度,对于构建高性能视频质量评价模型非常重要.我们在研究中发现,时域失真和时域波动分布是影响视觉兴趣性最重要的特征因素,因此定量度量视频序列的时域感知失真和时域失真波动,并根据这两个参量,采用自适应阈值判断的算法,标定出可能的兴趣像素点;同时对选中的可能兴趣点,进行空域连通分析.根据连通区域面积大小,确定1～5个可能的感兴趣区域.并根据聚类算法,确定最终的区域.最后结果证明了算法的可行性.

Video quality assessment models are significant for the perception of video coding.Different visual interests of the viewers were represented in different areas.The measurement of the level of interest of different regions was very important for building high-performance video quality evaluation models.We found that distributions of temporal distortion and temporal fluctuation were the most important feature affecting visual interest.According to the quantitative measurement of temporal distortion and temporal fluctuation,an adaptive threshold to calibrate the possible points of interesting pixels were proposed.One to five interesting regions were picked out with the spatial connectivity analysis of the selected pixels according to the final area.The results prove the feasibility of the algorithm.