Evaluation of Gabor Filter Parameters for Image Enhancement and Segmentation

Abstract

This work provides the information about the working of Gabor filters for different types of images. This comparative study is based on the output of the noisy and the filtered images using Gabor filter. Three types of Noises: Gaussian, Poisson and Speckle are used to produces Noisy images from noise free image for the purpose of evaluation. Frequency of Gabor filter and variance of Gabor filter is evaluated for approx. 200 images. These parameters are characterized on the basis of some standard image quality matrices like SNR, correlation coefficient and SSIM. The performance of Gabor filter is also evaluated by segmentation 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.