On Sampling Of Object Appearance Under Varying Light

Image-based sampling has been widely applied in many areas, including image pro-cessing, pattern recognition and so on. The appearance of an object may change sig-nificantly under different illumination conditions (such as the difference of the position of light source and properties of light). According to this rule, image-based sampling method samples the appearance of an objects by using changes of light source or view angle. Then these sampled images are used to represent the reflectance of this object. Image-based relighting is another method, which adds environment light map to the reflectance in-formation and gets new appearance of object that under a new condition. Therefore the quality of these base images is the key to image-based relighting. Currently, two important problems should be solved in the area of image-based sampling:first, how many images would be sufficient for rendering; second, how to sample the appearance of an object (the way we sample the appearance of an object). Since little is known about the appearance of object, it was a common practice is to use base images as possible. However, it costs too much time and space, which may even affects the realization of related rendering.To address two problems mentioned above, we introduce a new method to improve the result of relighting rendering by rebuilding new base image. Our method is based on spherical harmonic(SH) theory and bidirectional reflectance distribution function(BRDF) theory. This method can reduce the total number of base images and the time of sampling process. The main content includes studying and analyzing of SH and BRDF theory-based sampling method, HDR images and how to get the position and distribution of sampling points in spherical space. We focus on how to rebuild base image by using related knowledge and rebuild analysis for three kinds of BRDF Model. Experiments in my work contains two parts, one is sampling process of base images, the other is image-based relighting rendering. In the first part, we focus on sampling process and rebuilding base images. In the second part, we evaluate the affect of the result of relighting rendering. Besides, we do some experiment on HDR and get some good results. The final experiment shows that this method not only improves the rendering result, but also reduces the total number of base images and improves the quality of base images.