Enhancement of Medical Images Using Super Resolution Convolutional Neural Network

Abstract

The high contrast for images taken of a human body by the medical apparatuses is quite important to diagnose the patient case perfectly. In this paper, a strategy for enhancing the contrast of Computed Tomography (CT scan) and Magnetic Resonance Imaging (MRI) is suggested. The strategy consists of two stages: pre-processing and then enhancement, either using Gaussian blur or not. The pre-processing stage involves an image smoothing, convert the image color space from RGB to YCrCb. The brightness compound (Y) is implied withe the Supper Resolution Convolution Neural Network (SRCNN) in order to raise image resolution, and then the images are returned to color space RGB using transformation equations. The measures of Peak-Signal to Noise Ratio (PSNR), Structural Similarity Index (SSIM), Mean Square Error (MSE) and Universal Quality index (UQI) were used to assess the quality of enhanced images. According to the results, when a Gaussian filter is utilized, a higher resolution image is obtained, both in term of subjective or objective assessment. The use of a high-resolution convolutional neural network to enhancement medical images helps to detect tumors in early stages, and therefore the proposed system will help save human life.