Enhanced Image Colorization in Medical Image Analysis using Deep Learning Model

Abstract

Grayscale images are images that contain only intensity values. We use a process where we capture a grayscale image and give an image that is in semantic colours and tones of the input. This process is known as Image colouring (for example, the model colours it to hot pink). Image colouring can be used in many areas, including old black-and-white photography, old film, medical and scientific image colouring. Colouring is very important, but rewarding, because a colour image that looks natural must be taken from every grayscale input. Existing approaches to colouring black-and white images rely on manual human annotation. This often leads to incredibly desaturated results as true colouring. With less degree, the colour is more subdued (more black or white is added). Unlike traditional old-fashioned techniques, neural network-based colouring techniques are fully automated that do not require human assistance. From a variety of colouring techniques such as automatic colouring, semi-automatic colouring and hand colouring, convolutional neural networks and deep neural networks are chosen because they can process image classification and recognition datasets with high precision. In this paper, we are going to use the Convolutional Neural Network (CNN), a deep learning algorithm which is mostly used to analyse an image visually and Autoencoders, which reduces dimensionality of the input.