Empirical evaluation of image reconstruction techniques

Abstract

Image reconstruction is the process of manipulating an image to increase the amount of information perceived by a human eye. In this paper most popular filtering techniques have taken for comparison, that are Non Local Mean method, Particle filtering and Markov random fields. The Original NL Mean method replaces a noisy pixel by the weighted average of pixels with related surrounding neighbourhoods. In order to accelerate the algorithm; the filters are used to eliminate unrelated neighborhoods from the weighted average. The particle filtering technique will give statistical behavior of the image. The most appropriate window or neighborhood shape and size to estimate the image intensity in a given position. One attempt is to do perform filtering by selecting the neighboring pixels in a random fashion but without taking image structure into account. MRFs can be used as parametric models for the probability distribution of intensity levels in an image. The resulting framework explores optimally spatial dependencies between image content towards variable bandwidth image reconstruction. The results of techniques Non Local Mean method, Particle Filters and Markov random fields are compared by using two parameters such as PSNR and MSE values for the reconstructed images.Markov Random Fields method provides a better result when compare to Nonlocal mean method and Particle Filter.