A Survey on Image-Based Approaches of Synthesizing Objects

Synthesizing virtual objects into real scenes is an important process in augmented reality. Illumination consistency is the key problem of expressing the photorealistic rendering result. Consequently, a wide range of solutions for illumination estimation are proposed, such as capturing real scenes and estimation from single image. Because of its convenience, image based methods have become a popular way to synthesize virtual objects gradually. This paper provides a survey on state-of-art image-based methods for illumination estimation. As for a special case, outdoor illumination is discussed separately because of different assumptions and process. Through discussion of advantages and shortcoming of existing methods, main challenges and future work are explained at last.