A Novel Narrowband C-V Model for Saliency Object Extraction
Gang Feng Zheng, Yiwen Dou · 2018
In this paper, a unified frame of object extraction under complex background is proposed according to the empirical mode decomposition (EMD), saliency model and Snakes model. It aims to reduce the computational complexity of the object extraction under complex background. Firstly using EMD, a color image is decomposed to different Intrinsic Mode Functions (IMFs) which forms IMF sets. By biological inspiration mechanism, every IMF elements is extracted according to local visual saliency. Then, combining all local visual saliency, the global visual saliency whose edge is seem as the initial narrowband boundary of C-V model is formed. Lastly, the extracted object is obtained by curve evolution. The proposed method was evaluated using two benchmark databases GraKriWei and the motion images of our real parallel robot. The experimental results on both them demonstrate the effectiveness and robustness of the proposed algorithm.