While RGB is the status quo in machine vision, other color spaces offer higher utility in distinct visual tasks. Here, the authors have investigated the impact of color spaces on the encoding capacity of a visual system that is subject to information compression, specifically variational autoencoders (VAEs) with a bottleneck constraint. To this end, they propose a framework-color conversion-that allows a fair comparison of color spaces. They systematically investigated several ColourConvNets, i.e. VAEs with different input-output color spaces, e.g. from RGB to CIE L* a* b* (in total five color spaces were examined). Their evaluations demonstrate that, in comparison to the baseline network (whose input and output are RGB), ColourConvNets with a color-opponent output space produce higher quality images. This is also evident quantitatively: (i) in pixel-wise low-level metrics such as color difference (ΔE), peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM); and (ii) in high-level visual tasks such as image classification (on ImageNet dataset) and scene segmentation (on COCO dataset) where the global content of reconstruction matters. These findings offer a promising line of investigation for other applications of VAEs. Furthermore, they provide empirical evidence on the benefits of color-opponent representation in a complex visual system and why it might have emerged in the human brain.
The appearance of color stimuli with luminance levels beyond the diffuse white is gaining importance due to the popularity of high dynamic range (HDR) displays. Past work on color appearance of stimuli, color appearance models, and uniform color spaces mainly focused on the stimuli with luminance levels below the diffuse white, which were produced using surface color samples or conventional standard dynamic range (SDR) displays. In this study, we focused on the perception of white appearance for stimuli with luminance beyond the diffuse white. Human observers adjusted the color appearance of a stimulus to the whitest under different adapting conditions, including a dark condition and 12 illuminated conditions. It was found that the chromaticities for producing the white appearance under the dark condition were generally similar to those under the 6500 K conditions, regardless of the adapting luminance levels. In comparison to a recent study focusing on the stimuli with luminance below the diffuse white, the perception of white under the conditions with the adapting CCT levels of 2700, 3500, and 5000 K was significantly affected by the lightness level of the stimulus, which cannot be accurately characterized by CAM02-UCS. The results can be used for reproducing white appearance for highlights in HDR scenes. Further investigations on uniform color spaces for characterizing stimuli with luminance beyond the diffuse white are urgently needed for processing and displaying HDR images.