Perceptual Equivalence of Scale and Chromatic Aspects of Environmental Saliency Arising in Naturally Complex Scenes

A new framework is presented for saliency based analysis of naturally complex scenes. In this framework, the fluctuation of local scale shift and the diversity of subtle chromatic scattering are extracted as the environmental saliency: viewer specific visualization of sign patterns to be associated with landmark objects. Being guided by the probability distribution of the scale shift, the scene image is partitioned into a fractal attractor spanning the roadway area and associated distribution of boundary objects. The chromatic diversity of the scene image is evaluated within a probabilistic color space as the support of a saliency index. The saliency index is applied to the estimate of the invariant measure to capture the sign patterns as fractal attractors within the boundary distribution. Thus, the scale-chromatic aspects of the environmental saliency are articulated into a system of fractal attractors of perceptually equivalence to jointly identify ground-object structure. Through experimental studies, it has been demonstrated that the complexity of decision steps in the detection of landmark objects can be significantly reduced by the saliency based articulation. Adding to it, the perceptual equivalence between the scaleand chromatic-aspects of the environmental saliency is testable via the multi-fractal articulation; to this end, it is sufficient to match the fractal attractors designed in 2D chromatic aspect with 3D landmarks through the projection into 2.5D scale aspect. By this perceptual equivalence, a self-reflective mechanism is induced in the two-aspect representation of the environmental saliency.