基于形态学的视频序列人体骨架提取

Morphology- Based Human Skeleton Extraction in Sequence Images

本文在数学形态学算法的基础上,出了一种新的序列图像中人体目标骨架提取方法。首先运用数学形态学对二值图像的滤波、腐蚀和膨胀功能平滑图像中人体目标的边缘,去除背景中误分割出来的噪声。之后运用形态学算法中的击中击不中变换细化目标,提取骨架。运用单人和多人运动图像序列进行实验,结果证明,本文提出的方法效果较好。

The paper proposes a novel human skeleton extraction method in sequence images based on mathematical morphology. First- ly, the filter, erosion and dilation capacity in the morphology are used to smooth the edge of human objects and removes the noise being miss classified in the background. Then, the Hit- or- miss transform of the morphology is employed to thinner the human object so that to extract the skeleton. The experiment uses single and multi object image sequences to apply the method and the result show that our method is more efficient.