Performance measurement of various hybridized kernels for noise normalization and enhancement in high-resolution MR images

Abstract

In this article, a focus is laid on the hybridization of various noise removal kernels that are used in the normalization of the noise in the medical MR images which is acquainted into the images during the processes of image rendering caused due to inappropriate calibration of the equipment and poor illumination of the subject. In the process of statistical study of various kernels that include Otsu-based Adaptive Weighted Bilateral Kernel (AWBK), Adaptive Contourlet Transform (ACT), Adaptive Fuzzy Hexagonal Weighted Mean (AFHWM) Kernel, and Adaptive Multiscale Data Condensation Kernel (AMDC), the experimentation is carried over images that are corrupted at distinct noise levels. During the recovery of the noisy image, the performances of the various included approaches have been evaluated and presented in this article. Upon practical implementation, it is observed that each of those hybridized kernels outperformed the type of noise on which they are experimented. The mean computational time of each kernel is also been presented in the results.