一种基于纹理特征的教案视频镜头分割方法

该文分析了现有的对帧间特征差进行阈值比较的镜头分割方法,以及通过颜色空间和可调时间阈值进行视频聚类方法的不准确性,针对教案视频中大量文字内容体现出的特有的纹理特征,提出了使用基于灰度级共生矩阵纹理特征的C均值模糊聚类算法进行教案视频镜头分割.算法选取灰度级共生矩阵统计量之一的惯性矩作为度量帧间相似性的特征值,并根据教案视频中手写操作的特点调整特征向量,以此作为样本数据点进行模糊聚类.实验结果显示这种方法对于教案视频镜头的分割具有较好效果.

A Shot Segmentation Algorithm for Teaching Video Based on Texture Feature

This paper analyzes the shortage of shot segmentation technique based on the comparison between the inter - frame difference and the threshold, and the clustering method based on the color space and time threshold, considering the proper texture feature of mass characters in the teaching video, consequently presents a fuzzy clustering method based on the GLCM. The intertia quadrature, one of statistical values of GLCM, is selected as the character to measure the comparability of frames ; then, because of the handwritten operation the adjusted eigenvectors are as the samples for clustering. The experiment results prove the good effect iveness of this solution.