Multichannel Image Blind Deconvolution for Noisy Measurements

Abstract

This paper analyzes Multichannel Blind Deconvolution (MBD) techniques over noiseless and noisy data. We employ both REgularization by Denoising (RED) and the multichannel blind criterion as components of an approach to regularize the illposed inverse problem of image deblurring. Tests with different optimization techniques, such as fixed step, global search and the Barzilai-Browein step direction were performed in synthetic data and the results were compared with other techniques in the literature. Low Root Mean Squared Error (RMSE) kernels and image estimates were achieved in the deblurring of noisy images. This was achieved by combining fixed-step RED and the global search method in an alternating optimization algorithm.