A layer-based segmentation method for compound images

Segmentation is an important and necessary step to minimize the loss of text readability in compound image compression algorithms. This paper proposes a new MRC-based segmentation approach which computes the binary mask using morphological-based method applied on results obtained by Prewitt filter, Sauvola method. The method separates textual region from background using a layer filling algorithm. Experiment results show that compared to the Djvu segmentation method, a satisfying text-image separation could be achieved, and clear texts could be obtained, which confirms the effectiveness and robustness of the proposed compound image segmentation algorithm.