Fractal coding of image-color spaces for saliency-based object detection in naturally complex scenes

Abstract

A saliency-based approach is presented for object detection in naturally complex scenes. By regenerating the chromatic diversity in a probabilistic color space, the distribution of saliency colors is extracted as the viewer specific visualization of landmark objects. The saliency distribution is articulated into a system of fractal attractors spanning object images. Detected fractal models are visualized according to the perspective underlying the scene image.