Optical Remote Sensing Image Quality Assessment For Degraded And Restored Images

It is of great signigicance in many aspects of application of remote sensing technology to establish effective and stable remote sensing image quality assessment.This paper focuses on the research of quality assessment of single-distorted,multiply-distorted optical remote sensing images and restored blurred images.A remote sensing image quality assessment index based on the ratio of spatial feature weighted mutual information is proposed.The index RSIQA(Remote Sensing Image Quality Assessment)is calculated by the two kinds of mutual information,the spatial feature weighted mutual information between the source remote sensing image and the perceived image through the visual distortion channel,as well as the spatial feature weighted mutual information between the distorted remote sensing image and the perceived image through the visual distortion channel.A database ORSID(Optical Remote Sensing Image Database)is established.The subjective consistency of the proposed method was compared with other IQA methods on ORSID.Experimental results show that the proposed method has high degree of subjective and objective consistency,and high evaluating effectiveness for the quality of remote sensing images.A multiply-distorted remote sensing image quality assessment index based multi-scale deep feature similarity is proposed.The index MSDFSIM(Multi-Scale Deep Feature SIMilarity)is calculated by the two kinds of deep feature similarity,the deep feature similarity of the output maps between resized reference images and distorted images through a convolutional neural network model,as well as the deep feature similarity of the output maps between information blocks of reference images and distorted images through a convolutional neural network model.Then the two kinds of deep feature similarity are summed up to the quality assessment index on multiply-distorted images,MSDFSIM.A database MDORSID(Mutiply Distorted Optical Remote Sensing Image Database)is established.The subjective consistency of the proposed method was compared with other IQA methods on MDORSID.Experimental results show that the proposed method has high degree of subjective and objective consistency,and high evaluating effectiveness for the quality of multiply-distored remote sensing images.A restored blurred remote sensing image quality assessment index based on the joint assessment of a ringing metric and a sharpness metric.The index RBRSIQA(Restored Blurred Remote Sensing Image Quality Assessment)is calculated by the joint assessment of an ERM(Edge Ringing Metric),a BRM(Boundary Ringing Metric)and a SM(Sharpness Metric).The joint method is based on the nonlinear fitting using a BP((Back Propagation)neural network.A database RBORSID(Restored Blurred Optical Remote Sensing Image Database)is established.The subjective consistency of the proposed method was compared with other IQA methods on RBORSID.Experimental results show that the proposed method has high degree of subjective and objective consistency,and high evaluating effectiveness for the quality of restored blurred remote sensing images.