Hierarchical Saliency Detection Under Foggy Weather Fusing Spectral Residual and Phase Spectrum

It brings great difficulty for salient object detection on the road under foggy weather, owing to the low contrast and the low resolution of fog-degraded image. The traditional methods of saliency detection such as spectral residual approach and phase spectrum approach are out of work. According to this problem, a new hierarchical saliency detection approach based on transmission information of foggy images is proposed, which fuses the spectral residual and phase spectrum under transmission information with a new weighted fusing approach. The experiments show that the proposed method can detect salient object such as pedestrians and vehicles on the road more effectively than independent spectral residual and phase spectrum approach, even under heavy fog condition.