Context-based multiscale classification of document images using wavelet coefficient distributions

In this paper, an algorithm is developed for segmenting document images int o four classes: background, photograph, text, acid graph. Features used for classification are based on the distribution patterns of wavelet coefficie nts in high frequency bands. Two important attributes of the algorithm are its multiscale nature-it classifies an image at different resolutions adapt ively, enabling accurate classification at class boundaries as well as fast classification overall-and its use of accumulated context information for improving classification accuracy.