多方向自然场景文本检测

针对自然场景图像背景复杂和文本方向不确定的问题,提出一种多方向自然场景文本检测的方法。首先利用颜色增强的最大稳定极值区域(C-MSER)方法对图像中的字符候选区域进行提取,并利用启发式规则和LIBSVM分类器对非字符区域进行消除；然后设计位置颜色模型将被误滤除的字符找回,并利用字符区域中心进行拟合估计文本行倾斜角度；最后通过一个CNN分类器得到精确的结果。该算法在两个标准数据集上(ICDAR2011和ICDAR2013)上进行了测试,f-score分别为0.81和0.82,证明了该方法的有效性。

Multi-Orientation Natural Scene Text Detection

The background of text regions in natural scene images was complex, and the orientation of the text-line was uncertain, to address this problem, a multi-orientation natural scene text detection method was proposed. First the color-enhanced maximally stable extremal region(C-MSER)was used to create character candidates for natural scene images, the heuristic rules and the LIBSVM classifier were used to eliminate non-character regions; then a position-color model was designed to retrieve the false negative candidates, and the centers of the character regions were used for linear fitting to get the inclination angle of text-lines; finally, a CNN classifier was used to obtain the accurate results. The proposed method was evaluated on two standard benchmark datasets, including ICDAR2011 and ICDAR2013, and got the f-score of 0.81 and 0.82, which demonstrates the effectiveness of the proposed multi-orientation natural scene text detection method.