基于梯度增强的新闻字幕分割算法

News Video Text Segmentation Algorithm Based on Gradient Reinforcement

新闻字幕的分割在基于语义的新闻视频检索系统中具有重要的意义,为此提出一种基于梯度增强的新闻字幕分割算法.该算法使用图像多方向梯度的加权和代替图像的标准方差,通过各方向权值的调节加强某些方向的边缘信息,以提高分割效果.与一些经典的自适应阈值分割算法相比,该算法不仅能够保留大部分笔画,也能有效地减少断笔问题.基于光学文字识别的实验结果证明了文中算法的有效性.

In the news video retrieval system important part. In this paper, an improved based on the semantics, text segmentation is an text segmentation algorithm based on gradient reinforcement is proposed for news video. Compared with the classical adaptive approaches, the standard deviation of a text image is replaced with the weighted gradient sum of a text image. The weights of special directions can be adjusted to enhance the edge of texts and get better segmentation results; most of the strokes can be reserved. The effect of our algorithm can also be proved by the optical character recognition results.