基于像素邻域点信息的藏文图像细化算法研究  

Research on a Tibetan Image Refinement Algorithm Based on Adjacent Pixel Points' Information

作　　者：刘芳 张云洋[2] LIU Fang;ZHANG Yun-yang(Titetan Information Technology Center of Tibet University,Lasa 850000,China;Tibet University Library,Lasa 850000,China)

机构地区：[1]西藏大学藏文信息技术研究中心,西藏拉萨850000 [2]西藏大学图书馆,西藏拉萨850000

出　　处：《计算机技术与发展》2018年第4期21-24,共4页Computer Technology and Development

基　　金：国家自然科学基金(61165010)

摘　　要：细化是图像处理和模式识别系统中的一个重要过程,在图像分析和图像识别中应用广泛。只有把多像素的线条细化为单像素线条轮廓才能准确地进行字符的切分和文字的特征提取,对后续字符的分析和识别起着关键的作用。根据藏文字符的结构和书写特征,首先对藏文数字图像利用局部自适应方法进行二值化处理,再采用基于基线的滤波处理噪声方法进行去噪处理,以尽量简单直观地还原字符最原始的真实信息。在细化过程中,通过对某个像素点的八个邻域点的连接情况,在对照矩阵中查找对应矩阵项的值判断该点是否能删除,对藏文字符各点逐一进行判断和细化处理,最终得到文字的骨架。该算法在藏文字符数字图像细化实验中效果良好,正确率高,实用性强。Refinement is an important part in image processing and pattern recognition system,which has been widely used in image analysis and image recognition.Only by refinement of the multi-pixel lines into a single pixel line contour,the segmentation of characters and the feature extraction of the text can be carried out precisely,which plays a key role for subsequent analysis and recognition.In this paper,according to the structure and writing characteristics of Tibetan characters,the Tibetan digital image is processed in binarization by means of local adaptive method and then denoised by filtering method to deal with noise based on the baseline,in order to restore the original information of the characters as simple and intuitive as possible.In the refining,the refinement algorithm determines whether one pixel point can be deleted from its eight adjacent points'information.It judges Tibetan character's all points one by one,and refines them to produce the text's frame.In Tibetan character recognition experiments this refinement algorithm gets the positive results with satisfactory accuracy and strong practicability.

关 键 词：藏文 数字图像 二值化 去噪 细化 邻域点 对照矩阵 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]