ZY-3 Image Cloud Removal Based On Dictionary Learning Algorithm

The high resolution,multispectral and dynamic data acquisition of remote sensing images are playing a more and more important role in national construction.However,due to the interference of cloudy and rainy weather,there are always problems of information loss,low utilization rate and serious waste of resources in optical remote sensing image.Although there are many ways to achieve remote sensing image cloud removal,but the common methods usually require precise image registration,cloud detection,precise threshold selection steps,has the disadvantages of complicated process,complex operation,subjective intervention,not high degree of automation etc.With the improvement and development of compressed sensing theory in recent years,the sparse representation method based on dictionary learning has shown a remarkable effect in the field of image processing,which provides a new way for remote sensing image cloud removal.Aiming at the problem of removing clouds of remote sensing image by the dictionary learning algorithm,based on the cloud model of remote sensing image,this paper analyzes and compares the existing remote sensing image cloud removal methods,the image compression sensing theory and the dictionary learning algorithms,uses doublel dictionary K-SVD and block based restoration method for remote sensing image cloud removal.And then improve the method,paper add a specific sorting rule into the process of dictionary atoms training,so that each image dictionary has its own image properties while its atoms also have a similar arrangement order to reduce the interference between image differences.Thus,a new algorithm named AO-DL is proposed to remove the cloud of remote sensing image,which produces a good effect and has the advantages of reducing the requirements of cloud registration and detection,simple operation steps,high degree of automation etc.To illustrate the performance of the proposed method,experiments on two sets of data of multitemporal ZY-3 images at the same area are discussed.In this paper,the cloud removal effect of AO-DL algorithm is analyzed,its advantages and disadvantages are summarized,and the future of remote sensing image cloud removal based on dictionary learning algorithm is forecasted.