Analysing texture features from individual observer simulations

Abstract

We investigated the impact of simulated individual observer colour matching functions (CMFs) on computational texture features. We hypothesised that most humans perceive texture in a similar manner, hence a texture indicator that is the least dependent on individual physiology of human vision would be most likely a potential fit to serve as quantified visually perceived texture. To this end, the following strategy was implemented: hyper-spectral image textures were converted into XYZ images for individual observer CMFs, contrast sensitivity function (CSF) filtering was subsequently applied on the XYZ images for visual simulation. Two types of texture features were extracted from the filtered images. Finally, the difference between the texture features were analysed for observers with disparity in their CMFs.