一种非刚体目标的视频分割算法

视频分割是目标识别的关键技术。论文在对现有算法的分析,尤其是对空间分割中K均值算法和时间分割中的背景登记算法分析的基础上,改进了K均值算法,提出了一种基于时空结合的符合人眼对颜色变化敏感因素的分割算法。其基本思想是对视频序列分别做时间和空间分割,然后根据贝叶斯分类法对上述分割结果进行区域合并,即二次分类,从而得到分割结果。实验结果表明,论文提出的分割算法在保证算法效率的前提下,提高了分割的精度。

A Kind of Non-rigid-object Video Segmentation Algorithm

Video segmentation is one of the key technique of the object recognition.This paper improves the K-means algorithm and proposes an spatio-temporal algorithm according with the character of human vision system based on analysis the K-means algorithm and temporal segmentation which is called background registration algorithm.The basic idea is segment the video frame by space and temporal segmentation separately.And then classifies the result of them.The segmentation algorithm not only improves the accuracy of the result but also maintains the speed of the algorithm.