基于CNN的图片颜色处理.pdf

Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic Image Colorization with Simultaneous Classification Satoshi Iizuka Edgar Simo-Serra Hiroshi Ishikawa Waseda University (a) Colorado National Park, 1941 (b) Textile Mill, June 1937 (c) Berry Field, June 1909 (d) Hamilton, 1936 Figure 1: Colorization results of historical black and white photographs from the early 20th century. Our model can obtain realistic colorization results whether it is a picture of landscapes (a), man-made environments (b), human portraits (c), or close-ups (d). Photos taken by Ansel Adams (a) and Lewis Hines (b,c,d), and are taken from the US National Archives (Public Domain). Abstract 1 Introduction We present a novel technique to automatically colorize grayscale Traditional colorization requires significant user interaction images that combines both global priors and local image features. whether in the form of placing numerous color scribbles, looking Based on Convolutional Neural Networks, our deep network fea- at related images, or performing segmentation. In this paper, we tures a fusion layer that allows us to elegantly merge local informa- instead propose a fully automated data-driven approach for col- tion dependent on small image patches with global priors computed orization of grayscale images. Our approach uses a combination using the entire image. The entire framework, including the global of global image priors, which are extracted from the entire image, and l