Interactive Document Images Thresholding Segmentation Algorithm Based on Image Regions

In order to overcome the two problems, parameters selection in the local thresholding segmentation methods which have many parameters usually and discontinuous problem among neighbor regions in the segmentation results, an interactive document images thresholding segmentation algorithm based on image regions is proposed in this paper with the priori knowledge or experience from the users. Firstly, the presented method divides the image into several regions roughly. Secondly, it sorts all the image blocks according to their standard deviation values, which are taken as a measure to tell how much information of the background and object every block contains. Thirdly, the users input interactive information to separate all regions into three parts: blocks containing background or target only, blocks containing a small amount of background or target, and blocks containing distributing equilibrium background and target. Finally, the binarization of every block is conducted according to different criterion. Extensive experimental results show that the proposed scheme yields more promising binarization outcomes and also has better performance for document images under normal and inadequate illumination conditions, compared with global methods, simply thresholding approaches based on regions, and Chou's method. Moreover, the introduced approach is also effective for part of the non-text images.