针对复杂彩色图像中文本的特征,提出了利用一组级联弱分类器甄别文本区域的算法。首先根据文本块的边缘特征遴选出备选图像块,而后将其送入设计好的级联弱分类器,依次利用文本块特征、颜色信息、区域特征和字符笔画特征准则判断备选图像块是否包含文本。采用专用的文本定位比赛用图进行实验的结果表明,该方法运算简单,定位时间短,定位准确率可达到94.1%,召回率为85.9%。
A hierarchical classifier is proposed to discriminate text regions for the color images.The texture feature parameters of candidate image blocks obtained by the judgment of the edge feature are taken as the input of the hierarchical classifier.The characteristics of text blocks,color,region and strokes are used successively to judge whether some texts are in the candidate image blocks.Compared with others,the method is simpler and costs less time.For the images provided by the text location competition,the ...