Development of medical image enhancement algorithms using edge information based methods

Abstract

Medical image enhancement improves the quality and facilitatesdiagnosis This thesis investigates seven methods of medical image enhancement by exploiting useful edge information Since edges have higher perceptual importance the edge information based enhancement process is interesting However determination of edge information is not an easy job In newlineaddition enhancement methods have limitations such as i Less effective for medical images containing wide range of anisotropic and directional features ii Noise influence Two dimensional 2D high pass HP filters wavelet transforms WT directionlet transform DT and non sub sampled contourlet transform NSCT are used to obtain the edge information Their multiple scale representation helps to suppress the noise In this thesis methods are developed to enhance Computer Tomography CT Chest X ray Retinal Mammogram Ultrasound and Blood smear images Sharpening is a simple enhancement process that emphasize high frequency component of the original image Sensitivity to noise and limited newlinedirections are the major drawbacks of conventional high pass filters used in image sharpening enhancement To combat with these difficulties 2D isotropic Hyperbolic Secant Square HBSS high pass filter is developed to enhance the CT images and retinal images by sharpening This 2D nonseparable HBSS filters provides directional selectivity and less noise sensitivity The improved HP filter responses are used to sharp the CT and retinal images The improved performance of this method compared with common unsharp masking USM in various levels of noise conditions Structural similarity measure SSIM qualitatively evaluates their newlineperformances newline newline