复杂背景彩色图像中的文字分割

提出了一种纹理和连通域特征相结合的彩色文字分割方法。基于文字的边缘纹理特征,粗略分割出可能的文字区域。计算各区域内的颜色统计直方图,进行二类颜色聚类。分析文字连通域的几何特征,滤除非文字连通域。计算文字连通域像素点的垂直投影,估计文字的宽度和文字间隔,判断是否为文字排列,对粗分割的文字区域进行校验,确定文字区域的边框。通过自然场景下拍摄的100幅图像的文字分割实验,证明了该文字定位和分割方法的有效性。

Text segmentation in complex color images

An algorithm combined texture-based methods with connected component methods was proposed for text segmentation in color complex image.The algorithm detected candidate text blocks by using the edge features between character and background.Bi-color clustering based on the color histogram was performed.According to the character geometrical features,several components were discarded.Based on the vertical projection of the retained ones,the character width and the gap between two characters were estimated for the text verification,and the boundary of text blocks was adjusted.Experiments conducted on 100 images captured from natural scene demonstrate the validity of the proposed algorithm.