An approach to Color Gradient Vector Argument considering Data Uncertainty

Despite their shortcomings, gradient and second derivative operators remain popular as edge detectors because human beings usually do not show strong directional preferences when detecting edges or boundaries. However, in a colour image an edge can be given not only if the respective brightness values of two neighbouring pixels are significantly different, but also if their chromatic components are significantly different, even though their brightness values are very similar. So, it is important to consider changes in colour between neighbouring pixels for getting efficient colour-edge extraction algorithms. Unlike the gradient vector magnitude, the gradient vector argument has not been approximated. It has been in part due to the three dimensions used for defining colour feature. In this paper we present an approach to colour gradient vector argument approximation that considers the reliability indicators of the Smith’s Hue and Saturation components introduced in [13] for taking into account colour vagueness and variability problems on a HSI space