基于运动变化信息熵的视频序列时域亚采样技术

视频序列的时域亚采样对于低比特率信道上的视频传输和存储空间受限条件下的视频存储具有十分重要的意义。目前普遍采用的时域等间隔亚采样方法有时会造成视频序列中重要的运动变化信息的丢失。针对这一情况,本文提出了基于运动变化信息熵的视频序列时域亚采样技术,通过帧间运动场分析,利用运动场总体能量和运动补偿后的残差能量综合描述视频序列的运动变化信息,并根据信息熵最大化准则确定各个时域亚采样时刻点。大量的对比实验证明该方法明显优于时域等间隔亚采样方法,它较完整地反映了序列图像的运动变化过程,更有利于对视频内容的理解。

Temporal Sub-Sampling Technique of the Video Sequence Based on Motion and Changing Information Entropy

The temporal sub-sampling technique of the video sequence is very important for the video transmission in the low-bit-rate channel and the video storage in the limited storage space. In general the temporal sub-sampling by equal interval has been widely used in the video processing, but some motion and changing information will be lost sometimes. In this paper the temporal sub-sampling method of video sequence based on motion and changing information entropy has been put forward. Through the motion field analysis between successive frames, both the total energy of the motion field and the remainder energy after the motion compensation are used to describe the motion and changing information of the video frames. And the sub-sampling point in the time axis can be obtained by the rule of information entropy maximization. Many contrasting experiments have demonstrated that this method is obviously better than the temporal sub-sampling method by equal interval. By this method the changing process of the sequence is represented perfectly. It provides better perception and understanding for the video.