Color information in a model of saliency

Abstract

Bottom-up saliency models have been developed to predict the location of gaze according to the low level features of visual scenes, such as intensity, color, frequency and motion. We investigate in this paper the contribution of color features in computing the bottom-up saliency. We incorporated a chrominance pathway to a luminance-based model (Marat et al. [1]). We evaluated the performance of the model with and without chrominance pathway. We added an efficient multi-GPU implementation of the chrominance pathway to the parallel implementation of the luminance-based model proposed by Rahman et al. [2], preserving real time solution. Results show that color information improves the performance of the saliency model in predicting eye positions.