Abstract: Document images contain information in several dominant intensity levels. Detecting these dominant levels helps segmenting the regions in the image that need to be processed differently, if subsequently the image was to be recognized or printed. In this paper we propose a multithresholding algorithm to segment different dominant intensity levels in a document image based on intensity histogram and intensity transition histograms of the image. The method involves determining the number of dominant intensity levels in the image and computing the necessary number of thresholds that facilitate the segmentation. !15
展开▼
