Automatic colorization of black and white images using convolutional neural networks

Abstract

In recent years, image colorization has become increasingly common in video and image restoration. Automatic colorization approaches based on deep learning have recently shown to be very successful. Researchers are very excited to dig up a lot of knowledge from our history as technology advances rapidly. Still, most of the bits of data from the past are in black and white images, which are becoming a little inconvenient compared to current surroundings. Image colorization is the process of applying colours to grayscale images, which used to be a time-consuming and labor-intensive task involving a lot of human effort. The colorization of black and white images is a prevalent issue in the machine learning and computer vision communities. This method previously involved considerable user intervention, such as putting multiple color scribbles, looking at similar images, performing segmentation, and Previous ways for colorizing grayscale images relied on manual effort. The results were often undersaturated and unlikely to be true colorizations. We proposed a novel approach that uses Deep Learning and Convolutional Neural Networks (DLCNN) to color the picture from grayscale automatically. We have taken the image net dataset and convert all of the images from RGB to Lab colour space. We feed the network channel as a grayscale input, and the network will have to learn to predict the values of and channels. According to experimental results, the proposed DLCNN outperforms current methods on various performance indicators and achieves state-of-the-art efficiency.