Prediction Analysis of Noise Component Using Median-Based Filters Cascaded with Evolutionary Algorithms
A. Ramya, Sudha Rajesh, G. Karthick, Sudha Rajesh, G. Karthick · 2022
Denoising a digital image is the most investigated field in an image processing. The process of image denoising is to mutate the noise population in digital image and retain its original feature. Mainly the medical images such as mammogram and ultrasound images will get degraded because the system of these devices is composed of low-dose X-ray radiation. These interferences caused by the screening devices gradually reduce the interpretation of the constructive features. The noises that occur in these medical images are salt and pepper, Poisson, white Gaussian and speckle noise. Even a high radiation screening devices like magnetic resonance imaging (MRI) have shortcomings with the noise occurrence from the background surrounding, human interference-like breathing and movement of the body parts. While processing the medical image degraded with the noises, it should be carefully handled, because it contains many minute features. Usually, medical images undergo the processes like denoising, segmentation, restoration, reconstruction and classification, and recognition. Whatever the process it deals with, initially it should process with denoising stage, which is mandatory for medical images. Advance medical imaging devices may not require preprocessing, because of their superior technology and features. But mammogram and ultrasound-like systems need the preprocessing step. Efficient filters are needed for eliminating these noises from the medical image. These filters should not degrade the image features and fine details. There exist many conventional denoising filters applied to various applications, but working on with medical images is infrequent at research level. The novel optimized denoising framework for high-density impulse noise medical image is studied in this work. The denoising filter used in this work is median based and consists of two different stages of processing such as detection and filtering the noise candidate pixels. The performance of the median-based filter is increased by cascading the evolutionary algorithms such as artificial bee colony (ABC) and bacterial foraging optimization (BFO) techniques. The proposed optimized filters are compared with the existing filters to observe its performance in both subjective and objective views. The introduced new optimized framework gives less complexity in terms of computation and preserves the edges and fine information of medical image.