基于笔画宽度检测和改进块修复的自动英文文本移除算法

Automatic English text-removal algorithm based on stroke width detection and improved image block inpainting

为了提高英文文本移除后图像的修补效果,提出一种全自动文本移除算法。该算法基于骨架计算笔画宽度,并采用无监督连通区域分类实现文本区域检测与勾画。在文字提取后的缺损区域修复中,对传统基于样本块的修复算法进行了改进,依据边缘结构信息强弱确定边界块的合理修复顺序,采用两步搜索策略避免相似块误匹配,基于低秩矩阵填充技术提高填充区域的逼真度。实验结果表明,该算法能够准确定位文本区域,文本移除后的图像无明显篡改痕迹。

To improve the inpainting performance of image after English text removed,an automatic text removal algorithm is proposed.The text regions are detected and delineated using skeleton based stroke width computation and unsupervised connected region classification.After text removal,an improved sample block inpainting algorithm is proposed to repair the destroyed regions.The contributions in the inpaintiing stage are choosing the reasonable repairing orders of those boundary blocks based on structural characteristics,adopting two steps searching strategy to avoid mismatching of the similar blocks,and using low rank matrix completion to improve the fidelity of the repaired regions.The experimental results indicate that the proposed algorithm can find the letter edges accurately.The text removed images have no obvious visual traces of having been tampered.