关键词: 复杂背景, 汉字匹配, 快速鲁棒特征（SURF）算法, 灰度分级, 字符分割

Abstract: Aiming at the low matching accuracy of Chinese characters, an improved algorithm of SURF is presented. The algorithm is based on gradation character segmentation. Contrast of image is enhanced by using gray level segmentation, and then with the gray level classification tree, all pixels in the image are processed to the tree model. According to the gray level classification, the main node is determined. Grey level corresponding to the main node level is used in image segmentation. According to the particularity of Chinese characters, the rotation invariance of SURF algorithm is cancelled. Experimental results show that the improved algorithm can improve the matching accuracy effectively, especially for text image of poor quality.

Key words: complex background, Chinese characters matching, Speeded-Up Robust Features（SURF） algorithm, gray classification, character segmentation