The invention proposes a method and a device of determining a saliency map for an image.The proposed method uses a processing device for executing the steps of: selecting one of at least two different predetermined weight sets each associated with the predetermined scene category by determining to which of the predetermined scene categories a scene depicted in the image belongs, each predetermined weight sets comprising weights for color dependent subbands, splitting the image into color dependent frequency subbands and orientation subbands by splitting the image into color components and applying wavelet transformation to each color component, determining early feature maps for the subbands by extracting visual features from the wavelet transforms by a center-surround mechanism based on a Difference of Gaussian, weighting the early feature maps using the selected weight set and fusing the weighted feature maps.