Semantic White Balance: Semantic Color Constancy Using Convolutional Neural Network.

The goal of computational color constancy is to preserve the perceptivecolors of objects under different lighting conditions by removing the effect ofcolor casts caused by the scene's illumination. With the rapid development ofdeep learning based techniques, significant progress has been made in imagesemantic segmentation. In this work, we exploit the semantic informationtogether with the color and spatial information of the input image in order toremove color casts. We train a convolutional neural network (CNN) model thatlearns to estimate the illuminant color and gamma correction parameters basedon the semantic information of the given image. Experimental results show thatfeeding the CNN with the semantic information leads to a significantimprovement in the results by reducing the error by more than 40%.