Realistic Image Illumination System And Self-adaptive Sampling Application

Realistic image synthesis technology is a significant part in computer graphics,with the development of computer science, more creative global illuminationalgorithms for synthesizing more realistic image is required, and that effective test bedis the fundamental condition for the advancement in this field. Based on the researchof these years, we propose a Monte Carlo global illumination test bed for virtualenvironment — RenderStorm in the paper, which is of highly modular architecture,good expansibility and suitability to synthesize still and animated images. The mainmodules and their corresponding functions are described detailedly, and also how therendering process are carried on is illustrated. In order to synthesize realistic imagesrapidly, we improve on adaptive sampling algorithm which is a technique based oncore global illumination algorithm -ray tracing. Entropy theory, fuzzy math theoryand ray differential technique are applied in the research. We use tools RMS, NCD toindicate the difference between two images. The paper compare the images which aresynthesized with classic Contrast adaptive sampling, entropy based adaptive samplingwith images of random sampling technique. Renyi entropy and Tsallis entropy arefirstly used in this field. Through large numbers of test, we try to find out the law ofhow the entropy works in the sampling field. Comparison and analysis of the imagescan make us reach the conclusion that new adaptive sampling techniques can focussamples on important area, then can improve effect of the realistic images, and it'ssignificative to apply entropy (Shannon entropy, renyi entropy and Tsallis entropy) toadaptive sampling. In the end the paper gives out plan for later research.