Scanned maps processing using wavelet domain hidden Markov models

Abstract

This paper seeks to find ways to remove the unwanted information from the scanned GIS maps using wavelet domain hidden Markov models. WHMMs have proven to be a valuable tool for signal denoising, while they preserve the edges, so they can be used to remove the dithering effect that occurs during the printing process of the map. Linework data can be viewed as edges in the scanned map image. And, since WHMMs are well suited to images containing singularities (edges), they provide a good classifier for distinguishing between linework and elevation data (smoother areas in the image).