Fast fully data-driven image restoration by means of edge-preserving regularization

The fully data driven deconvolution of noisy images is a highly ill-posed p roblem, where the image, the blur and the noise parameters have to be simul taneously estimated from the data alone. Our approach is to exploit the inf ormation related to the image intensity edges both to improve the solution and to significantly reduce the computational costs. To detect reliable int ensity edges, the image is modeled through a coupled Markov Random Field wi th an explicit, binary and constrained line process. Following a fully Baye sian approach, the solution should be given by the joint maximization of a distribution of the image field, the data, the blur and model parameters. A first, significant reduction in computational complexity is obtained by de composing this joint maximization into a sequence of Maximum a posteriori a nd/or Maximum Likelihood estimations, to be performed alternately and itera tively. The presence of an explicit and binary line field is then exploited to reduce the computational cost of the usually very expensive model param eter estimation step. On this basis, we derive efficient and fast algorithm s along with procedures which are feasible and effective for real-time appl ications, where the real-time requirements are not too strict. Indeed, the structure of these algorithms are intrinsically parallel, and thus suitable for implementation on high-performance machines, or on specialized hardwar e and allows the computation time to be greatly reduced. The experimental r esults show that the method allows one to obtain good blur estimates even i n the presence of noise, without any need for smoothness assumptions on the blur coefficients, which would polarize the solution towards often unreali stic uniform blurs. (C) 2001 Academic Press.