A dynamical analysis on non-local autoregressive model and its application on image reconstruction

Abstract

Abstract Natural images are highly structured, which reflects strong spatial redundancy underlying its pixels. In this paper, a dynamical analysis of the non-local autoregressive (NAR) model is researched for the first time. We prove that under the condition when the spectrum radius of NAR matrix is less than 1, the original image as the global stable fixed point of function system referring to a modified NAR model can be iteratively recovered via attractor reconstruction from an arbitrary start point.