论文部分内容阅读
This paper presents an automatic grayscale image colorization method using convolutional neural network.Besides the gray target image,the user doesnt need to provide a reference color image nor manual guidance.First we train the convolutional neural network using residual connections based on the VGG-16 model.For colorization,a grayscale image is forwarded through the network and using the highest layer infers some color information,then it up-scales the color guess and adds in information from the next highest layer.Experimental results and user study on a large set of images demonstrate that our colorization method is competitive with previous state-of-the-art methods.