Study and Analysis of Different Types of Gabor Filters for Image Enhancement and Segmentation

Abstract

In most digital imaging applications, high resolution images are usually desired newlinefor later image processing and analysis. The motivation behind high image newlineresolution is due to its two principal application areas, firstly, it provides newlineimprovement of pictorial information for human interpretation; and secondly it newlinehelps in better image processing applications. Image resolution describes the newlinedetails contained in an image, the higher the resolution, and the more image newlinedetails. The central aim of Super-Resolution (SR) is to generate a higher newlineresolution image from lower resolution images. High resolution image offers a newlinehigh pixel density and thereby more details about the original scene. An image newlinewith improved resolution is desired in almost all of the applications to enhance newlinequalitative features and is reported to be achieved by Super Resolution Image newlineReconstruction (SRIR). newlineThe main objective of this thesis is to study and analysis of different types of newlineGabor filters for image enhancement and segmentation. This thesis presents newlinevarious experiments on various wavelet lifting scheme and various Gabor filters newlinelike traditional Gabor filter, fast Gabor filter and log Gabor filter for image newlineenhancement and segmentation. Experimental results show that log Gabor filters newlineperforms better than other filters for image enhancement. In addition, we newlineexperiment with various wavelet lifting schemes for image enhancement. Further, newlinewe propose a new super resolution image reconstruction scheme for image newlineenhancement and segmentation based on both wavelet lifting scheme and Gabor newlinefilter. Proposed approaches based on Gabor filter and wavelet lifting scheming newlineworks as follows. Initially, low resolution images of same scene which are newlineusually rotated, translated and blurred are taken to form a super resolution newlineimage. Further, image registration operation orients translated, scaled and rotated images in similar way to that of source image.