一种HEVC压缩域的运动目标检测方法

Moving Object Detection Method in HEVC Compressed Domain

运动目标检测是智能视频分析的一个重要环节,现有的检测方法主要是在像素域中进行处理,存在计算复杂度高、检测目标不完整等问题.提出一种HEVC压缩域的运动目标检测方法,利用HEVC在编码过程中产生的运动矢量、划分结构、编码模式等编码信息,首先对运动矢量进行预处理得到运动矢量幅值图,然后利用编码划分结构和编码模式在空间域上对运动矢量幅值进行滤波以及更新Intra编码块的运动矢量幅值,接着对运动矢量幅值图进行膨胀和时间域的滤波,最终得到运动目标.实验结果表明,本方法大幅地降低了计算复杂度,并具有良好的运动目标检测效果.

Moving object detection is the key part of intelligent video analysis. The current detection methods are mainly designed in the pixel domain,and there have some problems,such as high computational complexity and incomplete object detection. To solve these issues,this paper proposes a moving object detection method in HEVC compressed domain bying using encoding information such as motion vector,partition structure and encoding mode. First,the motion vectors are preprocessed to obtain the amplitude map of motion vector. And then the amplitude map is filtered based on the spatial domain,and the amplitudes in Intra coding blocks are updated by using the coding partition structure and the coding mode. After that,the amplitude map is expanded,and filtered in the time domain.Finally,the moving objects are detected by using the amplitude map. The experimental results show that the proposed method can reduce the computational complexity greatly and have a good effect on moving object detection.