【作者】 刘新星； 汪增福；

【Author】 Liu Xinxing and Wang Zengfu(Department of Automation,University of Science and Technology of China,Hefei 230027)

【机构】 中国科学技术大学自动化系；

【摘要】 本文提出了一种面向复杂彩色图像的基于连接体搜索和多纹理特征融合的字符提取方法。该方法首先对色彩空间进行量化以消减色彩,然后在量化后的彩色图像中进行边缘检测和二值化处理,最后在综合利用边缘图及彩色图像进行基于边缘的连接体搜索生成字符候选区域后,利用字符的纹理特征最终确定字符区。实验结果表明,本文提出的方法行之有效。它吸收了基于连接体的方法定位准确以及基于纹理的方法处理复杂场景更有效的优点,降低了计算的时间复杂度,具有较高的性能。更多还原

【Abstract】 In this paper, a text extraction method for the complex color image is proposed. First, we generate edge feature map based on the sub-color image which is created by color reduction algorithm. Candidate text regions are then generated by merging bounding blocks which are extracted by utilizing edge-based connected-component searching method taking edge-divided color information into consideration. At last, texture and stroke feature of text are used to eliminate the false candidates. Experimental results show this method can extract text including Chinese and English characters accurately. Because of the integration of connected-component and texture based methods, this method can deal with the complex image more efficiently while locating the text accurately.更多还原