Abstract: To use the human visual attention concept to maximum benefit in the domain of image processing and thus extract important or the most prominent object in a static image. The research up till now in this area has been limited to indicating just the Focus of Attention (FOA) and not the entire region of interest. The salient object detection problem for static images is used to formulate salient object detection as one which separates a salient object from the background. In this paper a bottom up approach to compute the saliency map analogous to one calculated by human brain. The mathematical model is prepared, based on study of anatomy of physiology of visual system primates. Low level features are extracted based on color, contrast and intensity. After normalization and linear combination, a master map or a saliency map is computed to represent the saliency of each image pixel. Finally, the image is segmented out from the background. About 30 images are selected from a large database of images from Microsoft Asia which contains a salient object or a distinctive foreground object.
Keywords: salient object, static image, FOA, normalization, segmentation.