Designing a Framework for Noise dependent Filter Selection Algorithm

Abstract

This work provides a framework to design an automatic filtering algorithm, which will work according to the noise present in the image. For this purpose, can be used about the working of the non linear geometric, diffusion, enhanced Frost, Lee, Homogeneous mass area, Wiener and Median filter for different of images. This comparative study is based on the output of the noisy and the filtered images using these filters. Three types of Noises: Gaussian, Poisson and Speckle are used to produces Noisy images from noise free image for the purpose of evaluation. Parameters are characterized on the basis of some standard image quality matrices like SNR, correlation coefficient, mean square error and peak signal to noise ratio. The performance of these filters is evaluated by comparative study of values of these parameters of noisy, filtered and original images. Based upon this analysis, results are compiled. These statistical metrics are also displayed graphically and they are compared for both the noisy and the filtered images.