Место издания:Tomsk Polytechnic University Tomsk, Russia
Первая страница:26
Последняя страница:31
Аннотация:In this work we propose way of combining ordinary and Markov chain Monte Carlo rendering techniques in image space. We used per-pixel mask to separate pixels (which we want to run Markov chain on) from the rest of the image. The mask was obtained from the ordinary Monte Carlo noise analysis. The proposed method was tested with combination of Multiplexed Metropolis Light Transport as Markov chain technique and two ordinary Monte Carlo rendering techniques - Instant Bidirectional Path Tracing and Light Tracing. As a result, our method allows us to get better accuracy in comparison to ordinary Monte Carlo and a better visual perception in comparison to Markov chain Monte Carlo techniques with the same rendering time.