基于连通区域的传真图像版面分割与分类算法

针对传真图像,为了提高版面分割与分类准确率,提高处理速度,以连通区域为处理元素,通过合理设定阈值,将水平和垂直相邻连通区域进行合并,快速准确地分割图像。并且将版面分割过程与分类过程相结合,根据连通区域的大小建立矩阵,提取能够表征区域信息的八维特征,然后使用BP神经网络将版面区域分为文字区域和非文字区域两类。实验中得到版面分割准确率为89.2%,版面分类准确率为94.22%。实验结果证明,该算法能够快速准确地对传真图像进行分割和分类,具有较强的实用价值。

Fax image layout analysis based on connect components

In order to improve accuracy of layout segmentation and classification of fax images,segment images by merging horizontally/vertically adjacent connect components blocks by setting threshold properly.And combined image classify process with image segment process,obtained matrix based on connect components.Then,abstracted the feature vector of eight dimensions for representing block information.And classified document images as two types with BP artificial neural networks.In experiments,the accuracy of layout segmentation and classification was 89.2% and 94.22%.Experimental results show that the method is effective and feasible in document images segmentation.