Chroma Component Generation of Gray Images Using Multi-Scale Convolutional Neural Network

In this paper, we solved the problem of the colorization algorithms using convolutional neural networks. There are existing algorithms predicting chroma components to generate colorization images, but the color of generated images is usually dim and the saturation of them is poor. Thus, we proposed to solve the colorization problem by a pyramid-alike multi-scale convolutional neural networks and convert the color space of image from RGB to HSV to predict the chroma components. Experiments indicate our algorithm can produce colorized images with more accurate color and higher saturation than the existing work.