Adaptive Sampling Based on GH-Distance for Realistic Image Synthesis

Realistic image synthesis technology is an important part in computer graphics. Monte Carlo based light simulation methods, such as Monte Carlo path tracing, can deal with complex lighting computations for complex scenes, in the field of realistic image synthesis. Unfortunately, if the samples taken for each pixel are not enough, the generated images have a lot of random noise. Adaptive sampling is attractive to reduce image noise. This paper proposes a new GH-distance based adaptive sampling algorithm. Experimental results show that the method can perform better than other similar ones.