视频图像中维吾尔文字的识别研究

Recognition of extracting Uyghur texts from videos images

提出了从复杂背景视频图像中提取文字并识别的一套算法,利用自适应迭代算法提取视频中维吾尔文字,针对维吾尔文字的一些特点,利用合适的预处理方法保留维吾尔文字中的各种点及特殊笔画,同时有效地消除了复杂背景带来的噪声。考虑维吾尔文字书写的特点,利用滑动窗口法提取文字特征避免了文字分割,将产生的特征向量输入到隐马尔可夫模型(Hidden Morkov Model)中进行训练和识别。

This paper presents a recognition system for automatically extracting text in video images with complex backgrounds,using self-adaptive iterative algorithm extracts Uyghur characters from video image.After preprocessing text images,the system needn’t segment texts into characters,instead,the system extracts statistical features from a window which is sliding text image.Then the resulting feature vectors are input to the Hidden Markov Models（HMM）for training and recognition.