Filtering of the results of Monte Carlo ray tracing using multiple images

This paper presents a method of filtering using several input images obtained by Monte Carlo ray tracing with small changes in the position of the virtual camera inside the scene. The limiting case with the absence of a camera shift, which actually degenerates into a distribution of independent samples to several Monte Carlo images, is considered separately. This approach does not impose any restrictions on the method of generating Monte Carlo samples and allows to take advantage of methods such as ray tracing on GPU and quasi-Monte Carlo. Unlike existing analogues, the proposed method takes into account the spread of Monte Carlo samples’ values for each virtual sub-pixel, which provides an opportunity for handling outliers and smoothing of aliasing effects while preserving details in the image.