An Object Extraction Method Based on Oscillatory Network

Based on locally excitatory globally inhibitory oscillator networks (LEGION), an object extraction method was proposed. The oscillation network can segment image and extracted objects by the scene and object feature. In this paper, the normalized difference water index (NDWI) was used as the feature of water body to segment remote sensing image. Because of the good coherence of spectrum property of water body, the NDWI of the water body would have strong couple weight on NDWI. The special spectrum property of water was also considered as a feature in this paper. Because the proposed method can extract object with their features instead of the pixel intensity, it was robust to noise and advantage to keep the detail information of object. The results of experiment validate the effective of the proposed method.