Salient region detection based on multi-resolution

Abstract

Human eyes can detect visual salient region easily, but computational modeling of this basic intelligent behavior still remains a challenge. Here this paper presents a salient region detection method based on multi-resolution which can highlight salient regions with well-defined boundaries of object. First, the original image is sub-sampled into three multi-resolution layers. Then for each layer the luminance and color salient features are extracted in frequency domain and the significant values are calculated by using invariant laws of Euclidean distance in Lab space. In order to remove noise and enhance the correlation among the vicinity pixels, the normal distribution function is used to specify the salient map in each layer. Finally, the final saliency map can be obtained by normalizing and merging the multi-resolution salient maps.