基于边缘特征分析和线性判断的文本帧检测算法

为保证视频文本检测作用于含有文本的视频帧中,提出一种基于边缘特征和分量线性的文本帧检测算法。算法首先去除模糊视频帧,通过融合颜色和梯度信息以及文本聚类的方法得到边缘图,利用启发式规则滤除非文本分量;然后根据视频文本分量质心位置及最小外接矩形大小规律构建候选文本行;并结合文本行内字符结构特征定义线性关系。最后对候选文本行进行线性判断,若有候选文本行符合线性关系则将视频帧标记为文本帧。实验结果表明所提文本帧检测算法比其他类似算法检测性能更好。

Text Frame Detection Algorithm Based on Edge Feature Analysis and Linearity Judgment

[Abstract] To ensure that the text detection and recognition process is being conducted only for frames that consist of text, a new approach to text frame detection in video based on edge feature and linearity judgment is put forward. Firstly the fuzzy video frames are removed, and the methods of fusing color and gradient information and text clustering are used to get the edge map, then false text components are eliminated by using heuristic rules. Next, candidate text lines are formed according to the centroid position and the minimum circumscribed rectangle of text components, and the linearity of the text components is determined by combining with the feature of text line character structure. Finally, the method undertakes a linearity check on candidate text lines. If one text lines in a frame satisfy the defined linearity condition, then the frame is considered as a text frame; otherwise it is considered as a non-text frame. Experimental results show that the proposed method is superior in terms of detection accuracy compared with other similar existing algorithms.