Abstract
It have been considered as a difficult issue how to optimize the presentation of unstructured visual data, such as natural scene imagery, while taking into account the diversity of color perception that includes color weakness and color blindness. We propose a novel methodology to analyze the inter-subject perceptual difference in terms of visual saliency. Our method quantifies and visualizes the information loss among saliency maps that reflect different visual properties, enhancing efficiency, objectivity, and flexibility of the designing procedure. The psychophysical experiment confirmed that the saliency-based model predictions were consistent to the subjective visual recognizably judged by human participants. We also report several examples of the system application to television programs.
