We present in this paper a new system to segment and label document images by combining statistical and multiscale view of different image components. Texture of text, halftone and images are characterized by modeling the distribution of the wavelet detail coefficients using a mixture of k Gaussians. Model parameters are then estimated using the expectation maximization (EM) algorithm. Using the proposed alogrithm, halftone areas were succefully differentiated from text regions
Mohamed N. Ahmed, Ahmed Eid, "Wavelet-Based Document Enhancement" in Proc. IS&T Int'l Conf. on Digital Printing Technologies (NIP22), 2006, pp 256 - 260, https://doi.org/10.2352/ISSN.2169-4451.2006.22.1.art00066_1