Memory efficient quadtree wavelet coding for compound images

Wavelet-based image coders generally perform well on natural images, which are typically characterized by slowly varying image intensities. Their performance suffers, however on compound images containing both text and image data. We modify a quadtree wavelet coder to perform well on text image data by treating text blocks differently from non-text blocks. We combine wavelet domain processing of non-text blocks with spatial domain processing of text blocks, and achieve improved performance over purely wavelet domain techniques for compound images.