An Approach to Text Segmentation from Color Document Images with Graphic Background

Abstract

Many researches focusing on segmenting text from various documents have been made in the past years. Most of them were devoted to dealing with monochrome documents. On the other hand, how to process color materials is still an open research field. Compared with monochrome documents, the processing of color documents will he much more complicated due to its wide variation in the color of text strings and background. In this paper, we present a procedure for separating text from color document images with graphic background. Our approach adopts the bottom-up scheme in processing. First a binary edge image is created for find out the connected components. Next, we select those components which look like characters or part of characters and then classify them according to their colors. Finally, we further group those character-like components which are in the same class and aligned horizontally, and then check if they are really characters. Experimental results have demonstrated the feasibility of the proposed approach.