Remote sensing image super-resolution based on POCS and out-of-core
Yong Li · 2010
Software-based image super-resolution techniques can be used to increase the spatial resolution of remote sensing images with low cost and fully utilized resolution data.However,some problems have to be solved to make super-resolution techniques practical,including how to automatically register images while maintaining high precision,how to efficiently and accurately reconstruct super-resolution images using only 2—3 images,and how to process very big images efficiently and robustly.In this paper,the traditional image super resolution method,projection onto convex sets(POCS),was improved with SIFT features used to guide image registration,with an hierarchical structure kd-tree used to accelerate feature matching,and with an out-of-core strategy proposed to achieve super-resolution of very big remote sensing images.Experimental results demonstrate that the method can efficiently reconstruct a super-resolution image with more detailed features using only 2 images,which shows remarkable improvement in image quality compared with original images.