A Biologically Plausible Model for Selective Color Image Processing

É. Dinet and A. Bartholin (France)


Color imaging, Saliency map, Visual attention, Selective processing


Efficient selective processing models can be derived from human visual attention mechanisms to deal with complex scenes with limited resources. The main idea is to analyze in details only the visual locations that attract attention. Models of visual attention mechanisms select regions of interest according to saliency criteria as performed by human observers. The purpose of this paper is to introduce a biologically inspired model of selective image processing. The proposed model provides a saliency map of the color input image. The saliency map locates the local visual attractors of the scene. It is computed by relating chromatic and achromatic features to local spatial arrangements in a competitive process. The first experimental results show that the output of the proposed model effectively predicts the positions of the few first fixation points of observers where computational resources should be directed for a selective processing.

Important Links:

Go Back