基于复杂背景的彩色图像中维吾尔文字切分

Uyghur Text Segmentation in Color Image with a Complex Background

沿着基线并具有大量附加部分书写是维吾尔文一大特点,这些特点使复杂背景的彩色图像中维吾尔文字行与字的切分和识别成为一个既困难又有趣的问题。本文首先对复杂彩色图像进行灰度化,其次将彩色图像转换为灰度化的边缘图像,再对图像进行局域二值化,然后进行区域检测和边缘调整,初步实现了图像中维吾尔文字行的定位,紧接着根据定位结果从图像中切分出文字行,统计切分后的文字行在水平和垂直方向上的像素累计情况,查找最佳切分点,分离出文字行中的字母独立形式或几个字母连成的连体字母段。实验结果表明,文字行的切分准确率达到96%,字切分准确率达到98%以上。

Writing along the baseline and with a large number of additions is one major characteristic of Uyghur, therefore the line and word segmentation and recognition of Uyghur in a complex color image becomes a difficult but interesting problem. In the paper, firstly,the complex color image is changed to be a gray-scale image. Secondly,the color image is changed to be a grayscale edge image and further a binary image of partial area. Then after regional detection and edge-adjusting, the line position of the Uyghur is located. According to the line position,the text line is obtained from the image. Finally,based on the segmentation result,the accumulation of pixels along the horizontal and vertical line is analyzed to find the best cut-off points and separate out the individual letter or the letter section contains several letters in a text line. Experimental results show that the line segmentation accuracy can reach 96% and the word segmentation accuracy can reach 98%.