Remote Sensing Salient Map Fusion Method Based on Gradient Optimization

Abstract

Target detection, change detection, and region of interest extraction are important research areas in remote sensing image processing. In order to reduce computational redundancy and improve image processing efficiency and accuracy, visual saliency models are widely used in the preprocessing stage of these fields. In this paper, a novel of remote sensing salient map fusion method based on gradient optimization is proposed. The local and global salient maps are obtained by wavelet transform and spectral residual method. The gradient salient map is solved by the maximum gradient optimization, and the fused salient map is reconstructed by Haar wavelet. The experimental results show that the fused salient map can combine the effective information of local and global saliency maps, and the detection accuracy is better than the global saliency map or the local saliency map, which has a better effect than the salient map fused by the simple method.