Pansharpening Based On Multiresolution Analysis And Injection Model

Remote sensing technology emerged in the middle of the 20 th century.This technology has developed rapidly in the field of space exploration based on aerial photography technology.It turns out,remote sensing systems are powerful tools for monitoring the earth’s surface and atmosphere on a local,regional,and even global scale.It can provide mapping and classification of important coverage areas,including land cover features such as vegetation,soil,water sources,and forests.Remote sensing images have been widely used due to the rapid development of satellite sensors.However,the existing remote sensing sensors cannot obtain images with high spatial resolution and high spectral resolution at the same time because of the limitation of sensor technology and the influence of other factors.With the increasing demand for high spatial resolution multispectral images in practical applications,researchers have proposed a large number of methods for remote sensing image fusion.Pansharpening is a technology that combines PAN images with high spatial resolution and MS images with low spatial resolution to obtain MS images with high spatial resolution.The key of the pansharpening method is to retain the spectral information of the multispectral image and enhance the basis of its spatial information,without causing spectral and spatial distortion.With the continuous development of pansharpening technology,it can be roughly divided into four categories: methods based on component replacement,methods based on multiresolution analysis,methods based on variational optimization and methods based on deep learning.Among them,the method based on multiresolution analysis is valued by researchers because of its good results.In this model,injection of high frequency and injection coefficients are the key to the technology.The only way to obtain excellent results is to inject appropriate high frequency details into the up-sampled multispectral image through appropriate injection coefficients.For this reason,based on the injection model and multiresolution analysis,this paper conducts a detailed study and in-depth exploration of the problems in pansharpening.Two pansharpening methods are proposed,which are based on multiresolution analysis and injection models.The main innovations of the paper are:1)A new pansharpening method is proposed,which is based on injection detail optimization.This method performs pan-sharpening by obtaining more precise highfrequency details.The proposed method includes two parts.First,the low-rank fuzzy fusion model is designed to fuse high-frequency details of panchromatic images and multispectral images.In this model,low-rank components and sparse components are obtained by decomposing the high frequencies of panchromatic images and multispectral images,and corresponding fusion rules are designed according to their characteristics,and then the fused high frequencies are obtained through inverse transformation.In the fusion process,part of the details of the panchromatic image is replaced by the multispectral image.Directly using the fused high frequency as the injected details may cause information redundancy or spatial distortion.To solve this problem,in the second part,an adaptive detail supplement model is proposed to further optimize the injected details.Based on the similarity and correlation between the fused high frequency and the original high frequency of the panchromatic image,the final injected details are obtained by supplementing the fused high frequency details.Experimental results show that the proposed algorithm is superior to the current advanced methods in maintaining spectral information and improving spatial details.2)A new pansharpening method is proposed,which is based on sharpening multispectral images and optimizing injection coefficients.There are some important details in the multispectral image,but the spatial resolution of the multispectral image is low,and the direct use of its high-frequency components will lead to poor spatial information of the final sharpening result.In response to this problem,this paper performs multi-level sharpening operations on multispectral images,and improves the spatial information of multispectral images step by step through the idea of grading.Then,the high-frequency components of the sharpened image and the panchromatic image are extracted respectively.Multispectral images have high similarities with panchromatic images,so their high-frequency components also have such characteristics.Based on the similarities and differences between the two highfrequency components,this paper designs adaptive weighting coefficients for fusion,and obtains the ideal injection high-frequency.In terms of injection coefficients,based on the spatial and spectral relationship between panchromatic images and multispectral images,the edge matrix and spectral ratio are calculated,and an improved calculation method for high-frequency injection coefficients is proposed.Experiments on degraded data sets and real data sets,subjective and objective results show that the proposed method is superior to other advanced pan-sharpening methods,which proves the effectiveness of this method.