Saliency pattern detection via simulated saccade in naturally complex scenes
Kohji Kamejima · Society of Instrument and Control Engineers of Japan · 2011
Joint stochastic-computational dynamics is introduced in image-color spaces for associative detection of saliency patterns from complex scene images. By regenerating the chromatic diversity by fractal attractor controlled by as-is primaries, saliency patterns are separated from noisy background. Resulted saliency patterns are efficiently detected via saccadic dynamics nondeterministically spanning boundary objects.