Relighting Technique Research Based On Harmonic Images

Image-based rendering is a cutting-edge technique in computer graphic, and a production by cross of image and signal processing, virtual reality and computer graphic. Image-based rendering makes real or real-like objects and environment reproduce in virtual environment, which raises sense of reality and immersion of users. Image-based relighting is an important research field in it and also a hot point in graphics. By capturing real objects and environment and synthesizing to render relighting results, image-based relighting light effect to the objects by environment realistically.This paper focuses on image-based relighting which is an important and practical method to apply relighting. There are three main steps in relighting: data-capturing, data-dealing and synthesis computation. Based on different methods to capture and deal with data there are different research directions. Two main methods are divided by purpose of capturing data. One is modeling-based relighting and another is image-based relighting. Modeling based relighting is to reconstruct mesh of objects and get reflection properties from it inversely, which is limited by precision and cost of surface reconstruction when objects'surface geometry and light is too complicated. Image-based relighting is to capture the objects under customized lights and calculate with basis function but need too much time to capture.By linking hot technique spherical harmonics and relighting together this paper proposes a method with discretely representing function defined on sphere by spherical harmonics to decrease the number and time of basis images need to be captured and reduce time and space of relighting computation.This paper first concludes relative research and development of image-based relighting and meaning, definition and classification.Then we descript briefly principle and meaning of spherical harmonics and also representation and application in relighting. Afterward we talk about capturing basis images and representing incident light and object reflection by treating spherical harmonics as basis function and then make relighting results.Finally relighting detail and process is optimized in detail and process, and computation is accelerated with parallel computation capability of GPU to make computation efficiency increased observably.The experiment results show that capturing basis images and relighting computation could be rapidly done with relighting procedure and optimized method proposed by this paper, and relighting results in same harmonic images could be got in real time. At the same time, we use the relighting based on harmonic images in a simple application.