Extraction of photographic area from document images

Abstract

In accordance with the diversification and efficiency promotion of visual communication, various new services have been desired recently. One such service is facsimile which can transmit both black‐and‐white pictures and photographic pictures. However, it is rather difficult to process these two kinds of pictures with a single algorithm. To fulfill the demand for high transmission efficiency and high quality of the reproduced pictures, it is desirable to segment the documents into two categories and process them separately. For this purpose, the extraction of the photographic area included in documents becomes a very important theme. In this paper an extraction algorithm of the photographic area from the document images using the characteristics of edge elements and background gray levels is proposed. The features of this algorithm are: 1( automatic setting of parameters adaptable for each picture is possible; 2) area segmentation corresponding to the various gray levels of backgrounds is possible; and 3 segmentation accuracy is rather high. In the experiment presented in this paper, 125 document images which consist of 512 × 512 pixels with 7‐bit gray levels have been used. The background gray levels of these images are 27, 36, 42, 57 and 77, and into these images photographic images, which occupy 50, 60, 70, 80 and 90 percent of the total area have been inserted. Experimental results show that in every image the photographic areas have been segmented successfully in the sense of practical use. For 93.8 percent of the images, the differences between the inserted areas and extracted areas are only 0 – 7 blocks (0 – 27 pixels), and the usefulness of this algorithm has been proved categorically.