Design of simple and efficient algorithms for improving the quality of digital images

Abstract

With an advent of Machine and Deep Learning algorithms, medical newlineimage diagnosis has got the new perception in terms of diagnosis and clinical newlinetreatment process. Regrettably, medical images are more susceptible to newlinecapture noises despite the peak in intelligent imaging techniques. However, newlinethe presence of noise in images degrades the both the diagnosis process and newlineclinical treatment process. The existing intelligent methods suffer from the newlinedeficiency of handling the diverse range of noise appears in the versatile newlinemedical images. To alleviate this challenge, the thesis proposes a novel deep newlinelearning network which learns from the substantial extent of noise in medical newlinedata samples. The proposed deep learning architecture exploits the advantages newlineof the capsule network which is used to extract correlation features and newlinecombines it with redefined residual features. Additionally, final stage of dense newlinelearning is replaced with the powerful extreme learning machines to achieve newlinethe better diagnosis rate even for the noisy and complex images. The newlineextensive experimentation has been conducted using different medical images newlineand various performances such as Peak-Signal-To-Noise Ratio (PSNR) and newlineStructural-Similarity-Index-Metrics (SSIM) are evaluated and compared with newlinethe other existing deep learning architectures. Additionally, comprehensive newlineanalysis of individual algorithms is conducted. The experimental results prove newlinethat the proposed model has outperformed the other existing algorithms by a newlinesubstantial margin and proved its supremacy over the other learning models. newlineHowever proposed model is required for improvisation of the real time newlineimages by exploiting the self-adaptive learning algorithms newline