基于模糊聚类和时域跟踪的视频分割

Video segmentation based on fuzzy clustering and temporal tracking

本文提出了一种基于模糊聚类和时域跟踪的视频分割方法。它把视频序列的连续7帧图像作为一组,对每组中间那一帧图像进行初始分割,对组内其他帧图像采用时域跟踪方法进行分割。初始分割采用了改进的多特征模糊聚类方法;时域跟踪是利用上一帧的分割结果通过运动估计向下一帧投影,作为下一帧的初始分割来完成的。

This paper proposes a video segmentation method based on fuzzy clustering and temporal tracking. We regard seven frames in a video squence as a group and conduct initial segmentation to the middle frame and then use temporal tracking to segment other frames in the group. We adpot improvised multi-features fuzzy clustering in initial segmentation and temporal tracking is performed by projecting the segmentation mask to the next frame as initial segmentation mask.