基于模板匹配的新闻图像字幕行切分算法

News Image Caption Line Segmentation Algorithm Based on Template Matching

针对新闻图像中水平字幕行的字符切分问题,为了克服已有基于单字符切分方法造成的字符分裂问题,利用字幕行中字符的分布规律构造了响应函数,从而将字符切分问题转变为响应函数的最优值问题,最后基于优化结果进行字符切分.该算法主要包括两部分:首先,利用垂直投影直方图确定单个字符的粗略宽度,并根据该值构造一个可变长模板;然后,构造模板响应函数,根据不同长度模板的最优响应函数值确定单个字符的左右边界位置;最后输出切分结果.实验结果表明,对于粘连/非粘连字符图像,该算法均能获得较好的实验结果.

The research on the character segmentation of the horizontal caption line in news images was made in this article. In order to overcome the character splitting problem caused by existing single character based segmentation methods,a response function was proposed based on character distribution. The character segmentation problem is converted into an optimal problem,and the character segmentation can be attained by turning to the optimal result. The algorithm mainly contains two parts： First,the rough width of a single character is determined based on the vertical projection histogram,which is utilized to construct a variable length template; Then,the template response function is constructed and the left /right boundary position of a single character is determined by the optimal value of the response function of different length templates; Last,output the segmentation results. Experimental results show that the proposed method can obtain satisfactory results for adhesion / non-adhesion character images.