Skip to Main Content
Human visual system can detect salient region fast and reliably, however, it is a big challenge to build a corresponding visual computing model. In this paper, a model of salient region detection based on local and regional features is presented. Firstly, the image is divided into 8 × 8 sub-blocks. Secondly, the local feature and regional feature of each sub-block are calculated. Local feature which reflects the sub-block's salient character locally is obtained through the calculation of color features in sub-block's 8 neighbors; regional feature which reflects the sub-block's character globally is calculated in the 15 × 15 block size areas which center on each sub-block. In order to avoid losing some detail features in the single scale, the saliency feature is extracted from multi-scale, and the salient sub-block is combined with the above features together as the block's salient value. Secondly, salient edge has been calculated by the contrast of the color values in 4 color channels. Finally, Salient map can be extracted by combining the salient feature and salient edge together. The experimental results show that our model can extract salient objects in images fast and exactly, which can be used in the areas of image retrieval, compression, and so on.
Machine Learning and Cybernetics (ICMLC), 2012 International Conference on (Volume:3 )
Date of Conference: 15-17 July 2012