Kaoru Imao

A segmentation method for composite text/graphics (halftone and continuous tone photographs) documents

When processing composite documents for digital copy machines and facsimile which contain a mixture of text, halftone and continuous tone photographs, ideally, the text portion can be separated from the graphics portion and more efficiently represented than the multi-bit pixel bitmap graphics representation. The technique for realizing this necessarily requires very precise image segmentation technology so that both the text and image portions can be correctly separated and efficiently managed. In this paper, a new image segmentation algorithm called SPACE is presented. The main idea behind the method for segmenting text/graphics composite documents is based on a peak-pixel method for detecting the halftone graphics and edge detection performed on a continuous black and white representation of the image. The results presented here are based on 64-level gray-scale, 400 dpi scanned images; but the segmentation method should also be applicable to color images and, as the image quality improves, the method becomes more effective.