International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-9 Issue-5, March 2020 1882 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: E2912039520 /2020©BEIESP DOI: 10.35940/ijitee.E2912.039520 Critical Findings on Restoration of Magnetic Resonance Images Abstract: The explosion of numerous medical images lead to the development of many different techniques to provide an accurate result. Although the signal to noise ratio (SNR), resolution and speed of magnetic resonance imaging technology have increased, still magnetic resonance images are affected by noise, contrast, and artifacts. To provide the image content or features relevant to diagnosis, contrast enhancement and reduction of noise with preservation of actual content should be carried out. The purpose of this paper is to present a critical review of different types of noises with an overview of diverse techniques for denoising and contrast enhancement for magnetic resonance images and discuss the advantages and limitations of these techniques with broad ideology. Keywords: Denoise, Contrast enhancement, Magnetic resonance imaging, Image filtering, Gaussian noise. I. INTRODUCTION Magnetic Resonance Imaging is widely used in medical images for detailed diagnosis for interpretation of tissues and other underlying features. In general, X-rays are the most frequently used imaging type for the visualization and interpretation of hard tissue such as bone, while modalities such as magnetic resonance imaging, computed tomography, and ultrasound are used for the visualization and analysis of soft tissues in the human body. Despite having many pros of each of these modalities, each also has its respective cons. Specifically, the diagnosis and monitoring of several diseases are limited by numerous factors such as the contrast, brightness, or noise in the image. These factors degrade the performance of computer-aided diagnosis (CAD) systems [1] or lead to the misapprehension of tissues by doctors. Currently, the development of CAD systems using imaging data obtained from medical imaging modalities and image/pattern recognition techniques is progressing rapidly, and researchers are using medical imaging datasets to develop accurate and efficient methods to meet the current need for the diagnosis and detection of disease. Most computer-aided diagnosis systems include preprocessing as a major step before segmenting and classifying the images. Noises, uneven edges, stubby contrast, and other artifacts make more difficult for further steps involved in the diagnosis. The diagnosis of any disease depends on the quality of the image and features to be used. For proper interpretation of different tissues and other underlying features, quality of contrast is necessary. Revised Manuscript Received on March 08, 2020. Sujeet More, Research Scholar, School of Computer Science and Engineering at Lovely Professional University, Punjab. Dr. Jimmy Singla Associate Professor, School of Computer Science and Engineering, Lovely Professional University, Punjab. If the image is less enhanced, the performance of the system gets degraded due to the quality of the image. If the image is over enhanced, then the system can’t distinguish the different features in the image. Though, many images acquired from magnetic resonance imaging exhibit an inadequate contrast that leads to weak tissue boundaries between different types of hard and soft tissues, contributing to difficulties with further quantifiable measurements like segmentation, classification, and analysis of the tissue structures. This leads to improper processing of further experimentation. Figure 1 shows different imaging types with their stubby contrast. The perceptible quality of magnetic resonance images plays a very crucial role in the diagnosis of disease and that can be degraded by noise which exists in the image due to the acquisition process [2]. Noise is affected by instruments, transmission media, and different types of radiations. The noise affects both computer-aided diagnostic system and the analysis on the disease like feature selection, segmentation, and classification. Noise reduction in images is still a challenge for many researchers because noise reduction introduces the blurring of the image. Fig 1: Contrast of different types of imaging. The first row consists of (a) and (b) Ultrasound and fundus image respectively. The second row consists of (c) and (d) X-ray and Fluorescent image respectively. The third row consists of (e) and (f) mammograms and MRI images respectively. Sujeet More, Jimmy Singla