结合小波插值与K-means方法提取彩色图像文字-《计算机技术与发展》

文章信息/Info

Title:: Extraction of Color Image Texts Combining Wavelet Interpolation and K-means

Author(s):: ZHANG Kai-ge; MIAO Yi; LEI Jian-kun; PENG Ke-liang; JIANG Mu-rong

Keywords:: text extraction and recogntion; wavelet interpolation; K-means method; BP neural network

摘要:: 彩色图像文字的提取和识别是人工智能与模式识别领域中的研究热点.现有方法在处理彩色图像中背景与文字相近或者字体较小的图像文字时会受到复杂背景的影响.为了解决彩色图像中文字与复杂背景颜色相近和模糊不清与排列不规则的文字提取与识别问题,文中采取人工选取文字区域进行小波插值放大的方法使图像清晰化,再利用 K-means方法对文字色彩进行聚类得到单背景彩色文字图像,然后通过二值化处理和文字分割后再结合BP神经网络进行文字识别.实验表明,该方法不仅能处理大多数彩色图像文字的提取,而且在识别低对比度彩色图像文字方面比目前常用的方法要好

Abstract:: The text extraction and recognition of color images is the research focus in the field of artificial intelligence and pattern recogni-tion. In order to solve the problems of text recognition that the characters in color images are too small to see clearly and the color of the characters is similar as the complex background or the characters are arranged irregularly,use the method which combines the wavelet in-terpolation and K-means cluster to extract the color texts. First,choose the region including the color texts,then employ the wavelet inter-polation to enlarge the color image,use K-means method to cluster the color to obtain a single background image,then change this image into the binary image and segment the text. At last,the trained BP neural network is used for the character recognition. Some experiments show that this method can process the most color text extraction easily,and more efficiently for recognizing the text with the lower con-trast and complex background