Abstract: Detecting a complete moving target from video image sequences is?a hot spot in computer vision research, this paper proposes a moving target detection algorithm based on the joint histogram. The algorithm uses the joint histogram to describe the similarity between adjacent frames, through blocking successively on two adjacent frames in video sequences, calculates the joint histogram of adjacent frames corresponding block and combines the structural similarity index to gradually remove the background block, moving target itself does not be handled in the algorithm, thus it can effectively avoid detected moving objects appear empty phenomenon, finally it gets a complete moving target. Numerical experiments show that the algorithm has better detection.

Key words: moving target detection, joint histogram, similarity index, wavelet decomposition

摘要： 从视频序列图像中检测完整的运动目标一直是计算机视觉研究的热点，基于联合直方图提出了一种运动目标检测算法。该算法利用联合直方图来描述相邻帧间的相似性，通过对视频序列中相邻两帧图像进行逐次分块，计算相邻帧对应分块的联合直方图并结合构造的相似性指标以逐步去除背景块，算法对运动目标本身不作处理，就能有效地避免检测到的运动目标出现空洞现象，最终得到了完整的运动目标。数值实验说明了该算法具有较好的检测性。

关键词: 运动目标检测, 联合直方图, 相似性指标, 小波分解