Research of global illumination algorithms rendering in glossy scene

In computer graphic （CG）, illumination rendering generated realistic effect at virtual scene is amazing. Not only plausible lighting effect is to show the relative position between of the objects, but also to reflect the material of visual appearance of the vir- tual objects. The diffuse-scene rendering reflectance credibility has gradually matured. Global illumination rendering method for the glossy material is still a challenge for the CG research. Because of the shiny materials is highly energy reflection between the com- plex light paths. Whether we trace glossy reflection paths, or use of one-reflection or multi-reflection approximate above complex il- lumination transmission is a difficult working. This paper we gather some commonly used global illumination algorithms recently year and its extension glossy scene improvements. And we introduce the limitation of classical algorithms rendering glossy scene and some extended solution. Finally, we will summarize the illumination rendering for specular scene, there are still some open prob- lems.

Automatic Key-Frame Extraction from Optical Motion Capture Data

Optical motion capture is an increasingly popular animation technique. In the last few years, plenty of methods have been proposed for key-frame extraction of motion capture data, and it is a common method to extract key-frame using quaternion. Here, one main difficulty is due to the fact that previous algorithms often need to manually set various parameters. In addition, it is problematic to predefine the appropriate threshold without knowing the data content. In this paper, we present a novel adaptive threshold-based extraction method. Key-frame can be found according to quaternion distance. We propose a simple and efficient algorithm to extract key-frame from a motion sequence based on adaptive threshold. It is convenient with no need to predefine parameters to meet certain compression ratio. Experimental results of many motion captures with different traits demonstrate good performance of the proposed algorithm. Our experiments show that one can typically cut down the process of extraction from several minutes to a couple of seconds.

Performance driven expression mapping based on segmented examples

We present a method that combines performance-driven method with segmented 3D blendshape models to animate a face. First we prepare key sample examples and corresponding key target examples. Next we segment the whole face into two regions, for each region we reduce dimensionality of source examples using PAC into abstract space which is defined by truncated PCA eigen- vectors. Then for each example we fix the cardinal base function, which can determine the weight of the target example. Finally, in the animation stage we compute the weight of each example for each frame and add the weighted displacement vectors of each re- gion on the general face model.