Perception-inspired Tone Mapping

Abstract

The display of high dynamic range images and video requires a tonemapping algorithm to depict their original appearance on existing display devices whose capabilities in terms of dynamic range are insufficient. The insightful application of knowledge about human visual system can assure high fidelity of depiction in such an algorithm. In this thesis, we design new tone mapping models and improve existing algorithms by an informed use of human perception to provide a high fidelity depiction of high dynamic range. We develop a real-time tone mapping solution which reproduces the subjective appearance of dynamic HDR contents by accounting for perceptual effects that significantly contribute to the appearance of natural scenes. We design a computational model of lightness perception that can be applied to high quality tone mapping for static images to reproduce their original HDR appearance in terms of lightness. We identify common distortions typical to tone mapping which may hinder the comprehension of image contents, we design appropriate metrics to measure the perceived magnitude of these distortions and evaluate existing tone mapping algorithms accordingly. To compensate for observed distortions, we introduce a method which improves the tone mapping results beyond numerically optimized solution by using techniques strongly based on perception of contrasts. Presented solutions can be efficiently integrated in varied HDR applications including photography, playback of HDR video, image synthesis, light simulations, predictive rendering, and computer games.</p>