Blind Image Restoration and Segmentation via Adaptive Norm Regularization and Mumford-Shah Edge Indicator

This paper proposed a new model which can successfully solve the image restoration problem and segmentation together. In the new model, the blind image restoration and segmentation problem are decoupled, making the algorithm faster. The blind image restoration process is based on the adaptive norm regularization which can estimate both piecewise constant point spread function and smooth point spread function properly. The segmentation process is based on the explicit edge indicator function which is deduced by the Mumford-Shah model. The new model can be solved under Split Bregman iteration framework to accelerate the computation. Numerical experiments show that new algorithm produces the promising result and robust to noise.