Blind recognition of mixed and disturbed correlated ancient texts

摘要

Separation and recognition of ancient documents and texts that have been mixed and disturbed over centuries, is an interesting problem in image processing area and it has been investigated by many researchers. In recent years, independent component analysis (ICA) method has been used for solving this problem, but independence of sources is an essential assumption in ICA, whereas in some problems, sources are not independent. So in this paper, we have tried to propose a method for separation and recognition of mixed and disturbed correlated ancient documents and texts utilizing MUltiple Signal Classification (MUSIC) and Estimation of Signal Parameters via Rotational Invariance Technique (ESPRIT) algorithms from blind source separation (BSS) techniques. The good performance of this method has been investigated for real images.