基于数学形态学的渐变色字符定位算法

摘要

In order to deal with the character location of gradient background image ,a new method based on mathemati‐cal morphology is proposed .The method uses the inherent characteristics of characters efficiently .In this algorithm ,the edge extraction and expansion operation are used to obtain the whole character skeleton .And then the gradient background character is smoothed by the filling and corrosion operation .The top‐hat filter and the color‐inverted operation are used to brighten and carry out the whole extraction .Character extraction rate can be raised to 90% .Experiments showed that the proposed algorithm had a strong ability to locate characters on gradient color images and could effectively reduce most of the noises ,compared to the edge segmentation and neural network algorithm .%为处理渐变色字符的字符定位，文章提出了一种基于数学形态学的字符定位算法。该算法很好地利用字符的特有属性，采用形态学中的边缘提取与膨胀操作，取得整体的字符骨架，再通过填充与腐蚀操作使得渐变色字符周围变得平滑，最后利用反色操作与 top‐hat 滤波进行整体提亮，从而提取出字符。实验表明，该算法对渐变色字符有较强的定位能力，并能有效地去掉绝大部分噪声，字符提取率高达90％，与目前基于边缘分割，神经网络等算法相比有较大改进。