An approach to manage Hue and Saturation uncertainty addressed to colour segmentation algorithms

Variability of Hue and Saturation components is an important drawback for developing accurate colour image segmentation algorithms based on HSI spaces: input RGB noise gets unevenly amplified through RGB-to-HSI formulation. Therefore, some H S values are intrinsically more uncertain than others. To quantify such uncertainty, we propose to use our Hue and Saturation reliability functions, which are defined on predicted standard deviations of colour samples. These functions provide a reliability degree of the H S values of image pixels, so that any segmentation algorithm may ponder the relevance of input colour data.