Extracting salient visual attention regions by color contrast and wavelet transformation

Visual attention detection is an important technique in many computer vision applications. In this paper, we propose an algorithm to extract a salient object from an image using bottom-up and top-down computations. In bottom-up computation, segment-based color contrast and attention values are employed to compose a bottom-up saliency map. In top-down computation, in-focus areas of the image are extracted to derive attention values using wavelet transforms for constructing a segment-based top-down saliency map. Attention values from both maps are combined by linear combination. The foreground/background-based salient object extraction is applied to form an output object. Experiments on 1,200 color images show that the proposed algorithm yields high level of satisfaction.