Extracting Valuable Information Hidden in Large-scale Image Data for Novel Image Synthesis
Finished by March 31, 2014

Computer vision is a key technology to link two-dimensional images and three-dimensional real world, since many various images can be accessed readily in recent years. By integrating both physics-based and exemplar-based algorithms, we shall develop techniques to extract various properties of things seen in images using unorganized image dataset such as a massive amount of images available on the internet. Moreover, we shall apply our techniques for recovering reflectance models, image synthesis and image enhancement, which are usually very difficult, if not impossible.