Optimization driven deep learning Classifier for image enhancement And deblurring of medical images A THESIS

Abstract

Medical imaging has gained more attention in proper newlineexamination and diagnosis of physiological aspect of humans. Clinical experts newlinedepend on the image results obtained through image sensors. Every time, newlineimage collected by the device or sensor has great exposure time results in newlineblurred images. It is very difficult to examine the physiology and to obtain a newlinebetter diagnostic result with the blurred medical images. Image deblurring is a newlinemajor issue in the imaging domains that requires more consideration for newlinerecovering the essential structures of the images. The real-time images are newlinemostly blurred due to imaging devices that result in certain blurring level, and newlinethe image storage, transmission as well as compression also leads the image newlineblurring. Due to the insufficient information regarding the point spread newlinefunction, the image deblurring mechanism cannot provide better matching newlineresults. The main role of image deblurring is to pick up the clear images from newlineoriginal image that consists of blur because of the relative actions amongst newlinescene and camera, or the atmospheric effects in imaging procedure. To solve newlinethe image deblurring issues and to provide better enhancement in the imaging newlinescenario of a medical image processing system, it is more essential to develop newlinethe image deblurring and the image enhancement method. The major newlinecontribution of the research is to design the method for solving the issues of newlineimage deblurring and to generate better image enhancement results with newlinemedical images. The first contribution of the research is to design a method newlinenamed Ant Cuckoo Search Optimization (ACSO) based Image Enhancement newlineConditional Generative Adversarial Network (IE-CGAN) to perform the newlineIv process of image contrast enhancement by considering the medical images. At newlinefirst, noisy pixels are detected from original image, and the noise is removed newlineand thereafter image enhancement process is carried out using the deep newlinelearning classifier newline newline