How color profile affects the visual quality in light field rendering and novel view synthesis

Abstract

This paper investigates the impact of image color profile on the visual quality in light field rendering. Light field rendering methods require several input views, capturing the same scene from different camera positions. These views can be captured in real life by cameras or rendered from synthetic 3D scenes. A novel view can be synthesized from the input views, capturing the scene from a virtual camera position. Modern cinematography makes use of various color profiles that affect how the measured light on the camera sensor is translated into the color values of the digital image. Different profiles can be used for different purposes, be it artistic or pragmatic to reveal necessary details. The main scientific question of this paper is if certain color profiles can lead to better quality of the novel view synthesis in light field rendering. The paper shows that logarithmic profiles can significantly improve the visual quality. Three light field rendering methods from different categories were used to compare 17 widely used color profiles. Additional measurements show that post processing and color grading needs to be applied at the very end of the rendering pipeline to ensure the best quality. Also, the effect of denoising algorithms on light field rendering is measured.