Image Colorization: A Convolutional Network Approach

Abstract

In recent days, due to some application of image colorization as automatic coloring of old pictures, correctly identification of thief from sketch image, etc., image colorization has become hottest area in research domain. Machine learning plays an essential role in this concern. This paper illustrates deep convolutional neural network approach with VGG16 pre-trained classifier that takes grayscale image as input image and gives its equivalent-colored image as output image. In this research, deep convolutional neural network is categorized into encoder, fusion, and decoder part, and VGG16 pre-trained model is used for extracting high-level features from an image. The proposed network is trained till 2000 epochs. The final-colored image is compared with ground truth image.