中文摘要：

字符特征提取是含文字碎片图像拼接的关键环节，直接影响拼接效果。提出了一种包含文字信息的碎片图像特征提取方法，首先，对原始碎片图像进行预处理，分别得到字符和碎片背景的二值图像；然后，利用边缘检测算子提取字符和碎片背景的单像素边缘；最后，利用结构算子检测碎片图像中字符与碎片背景边缘的交叉点，并根据交叉点的信息提取字体边缘方向特征。仿真实验结果表明本文提出的碎片图像字符特征提取方法思路合理，能够快速、准确地提取出碎片图像中的字符特征，为后续的匹配拼接奠定基础。

英文摘要：

Feature extraction is the key of scrap paper image mosaics including character, which affecting the effect of mosaics. A method of character feature extraction of scrap paper image is proposed. Firstly, the original scrap paper images are preprocessed to get the character binary image and the scrap paper background binary image respectively. Secondly, the single pixel edges of character and scrap paper background are extracted by using edge detection operator. Finally, the crossing point of character and the edge of scrap paper background is detected by using structure operator, and the direction feature of character edge is also extracted according to the crossing point information. The simulating experimental results show that the character feature extraction method proposed in this paper is reasonable, which can extract the character feature of scrap paper image quickly and correctly, laying a foundation for the subsequent image mosaics.