Adaptive Algorithm For Remote Sensing Image Fusion And Its Application

Remote sensing is an important means to monitor the dynam-ics of the earth surface.It is still challenging for single-sensor systems to provide spatially high resolution images with high revisit frequency be-cause of the technological limitations.Spatiotemporal fusion is an effective approach to obtain remote sensing images high in both spatial and tempo-ral resolutions.Single image super-resolution aims to generate a visually pleasing high-resolution image from its degraded low-resolution measure-ment.It is used in various computer vision tasks,such as remote sensing image fusion,security and surveillance imaging,and medical imaging.Based on dictionary learning and sparse representation theory,this dis-sertation proposes the end-to-end fusion model and the adaptive multi-band constraints fusion model.Based on neural networks approximation theory,this dissertation improves the single image super-resolution method based on deep learning.The specific content of the work is described as follows:1.The conventional dictionary learning fusion models are usually com-prised of two steps:learned the dictionary pair in the dictionary pair train-ing step,and then used it in the reconstruction step.Those dictionary pairs are insufficient because of lacking the information in target time,especially the information of known low resolution image.To improve the represen-tation capability of dictionary pair,we propose an end-to-end fusion model which integrates the dictionary pair training step and reconstruction step into one single model.We use one-by-one update strategy to solve the model and update the dictionary pair.Inspired by semi-coupled dictionary learning method,our model also uses mapping function between high and low resolution coefficients.Compared with the conventional fusion model,the experimental results show our algorithm is better in metrics and visual effects.2.Though dictionary learning fusion methods appear to be promising for spatiotemporal fusion,they do not consider the structure similarity be-tween spectral bands in the fusion task.To capitalize on the significance of this feature,a novel fusion model,named the adaptive multi-band con-straints fusion model?AMCFM?,is formulated to produce better fusion images in this paper.This model considers structure similarity between spectral bands and uses the edge information to improve the fusion result-s by adopting adaptive multi-band constraints.Moreover,to address the shortcomings of the?₁norm which only considers the sparsity structure of dictionaries,our model uses the nuclear norm which balances sparsity and correlation by producing an appropriate coefficient in the reconstruc-tion step.We perform experiments on real-life images to substantiate our conceptual augments.In the empirical study,the near-infrared?NIR?,red and green bands of Landsat Enhanced Thematic Mapper Plus?ETM+?and Moderate Resolution Imaging Spectroradiometer?MODIS?are fused and the prediction accuracy is assessed by both metrics and visual effects.The experiments show that our proposed method performs better than state-of-the-art methods.3.”Zero-Shot”Super-Resolution?ZSSR?is an algorithm which ex-ploits the internal recurrence of information inside a single image,and train an image-specific CNN at test time,on examples extracted solely from the input image itself.To overcome the shortage of ZSSR,we propose the En-hanced”Zero-Shot”Super-Resolution?EZSSR?which uses an alternative method in downsampling step to keep more high frequency information.Moreover,EZSSR integrates dense network with ZSSR by re-designing the network structure,which improves its capability of feature extraction.Experimental results indicate that the EZSSR not only has good reconstruc-tion effect,but also enhances the algorithm stability.