Home // SIGNAL 2024, The Ninth International Conference on Advances in Signal, Image and Video Processing // View article

Mixture Based Hybrid Regularization Method for Blind Image Deconvolution

Authors:
Linghai Kong
Suhua Wei

Keywords: Poisson blind image deconvolution, Square Cauchy distribution, Multiconvex

Abstract:
Following our recent work on mixed Poisson-White Spike noisy image restoration, we present a multi-convex optimization model to address the fundamental problem of Poisson blind image deconvolution (BID). This problem is encountered in a special application of X-ray radiography in hydro-tests, which also plays an important role in advanced tomographic imaging. We utilize a combined two-dimensional Square CauchyGaussian distribution, whose parameters are totally unknown, to characterize the base structure of the convolution kernel. A new prior density function for the convolution kernel is proposed by integrating the structure density into a Kullback-Leibler divergence. The multi-convex optimization model is derived by a joint maximum a posteriori estimation (MAP) procedure, into which local estimation and expectation maximization algorithm are involved to gain convexity and solvability. To solve the proposed model numerically, a block coefficient descent based algorithm is to be proposed, in which majorization-minimization algorithm and Barzilai-Borwein estimation along with alternating direction minimization of multipliers are utilized to promote the computational efficiency. Numerical results show the effectiveness of our proposed algorithm, as well as its adaptivity.

Pages: 1 to 3

Copyright: Copyright (c) IARIA, 2024

Publication date: March 10, 2024

Published in: conference

ISSN: 2519-8432

ISBN: 978-1-68558-142-8

Location: Athens, Greece

Dates: from March 10, 2024 to March 14, 2024