In this paper, we analyze the applicability of Constant Modulus Algorithm (CMA), one of the most widely used and tested blind equalization technique to blind image deconvolution. With a detailed mathematical analysis, we show that the strong correlation between the neighboring spatial locations found in natural images becomes a major constraint on the convergence of CMA. In order to overcome this constraint, we introduce a novel image pixel correlation model in relation with natural image statistics. Based on this model, a segmented blind image deconvolution through CMA is proposed. The robustness of the proposed algorithm with natural images is discussed in terms of efficiency and effectiveness.
|Published - 2010
|International Conference on Signal Processing and Communication Systems (ICSPCS 2010) - Gold Coast Australia
Duration: 1 Jan 2010 → …
|International Conference on Signal Processing and Communication Systems (ICSPCS 2010)
|1/01/10 → …