Research On Remote Sensing Satellite Fusion Image Quality Assessment

With the continuous development of remote sensing technology,remote sensing technology has been widely used in agriculture,water conservancy,environmental monitoring and land cover.As the speed of remote sensing image data collection accelerates and the data update cycle becomes shorter,how to quickly screen effective data and build an effective image quality assessment mechanism is of great significance to the fields of security,remote sensing,and military.In addition,high spatial resolution panchromatic(PAN)images and low spatial resolution multi-spectral(MS)images acquired by remote sensing satellite sensors need to be fused with the fusion algorithm to make the image information richer and facilitate the later image information collection and processing.The key issue of fusion is the issue of its quality assessment.At present,there is much controversy about the design of objective assessment methods for fused images.Therefore,this thesis researches two aspects of panchromatic stereo images in the field of building detection and PAN/MS image fusion in the field of visualization,and establishes an effective objective assessment method.This thesis focuses on the following two aspects:(1)For the panchromatic stereo images of remote sensing satellite,this thesis proposes two objective assessment methods.a)A full reference quality assessment of satellite panchromatic stereo images based sparse representation is proposed.Firstly,a new satellite panchromatic stereo image database is constructed,which is different from subjective scoring.It uses the building detection accuracy index to represent the degree of image distortion.Then,an objective evaluation model is proposed,which extracts the features of the original image and the distorted image and performs dictionary learning.Next,four quality scores are constructed based on the similarity measure of the sparse representation.Finally,the final objective evaluation value is obtained through support vector regression.Testing on the constructed database,the experimental results show that the proposed objective assessment method has better performance.b)Based on the database for further research,a no reference quality assessment of panchromatic stereo images based joint dictionary is proposed.Firstly,the local and texture features of the training set images are extracted and combined with dictionary detection for detection accuracy.Then,use the same method to extract the features of the testing image and perform sparse representation with the trained dictionary.Finally,the objective prediction value of testing image is obtained.Testing on the constructed database,the experimental results show that the performance of the proposed objective assessment method is significantly improved.(2)For the fused images from fusion of PAN and MS images,this thesis is divided into two parts: verifying the performance of existing assessment methods and proposing a new objective assessment method.a)A large-scale database is proposed for the assessment of fused image quality.Firstly,a large fused image database is constructed,which contains 2270 pairs of original PAN and MS images and 13620 fused images,and is divided into four thematic types of Building,Green Vegetation,Water Scenario and Mixed Surface.Then,subjective testing is performed according to the visualized application of the fused image.Finally,the performance of the commonly used objective assessment methods is verified and analyzed.b)Based on the constructed fused image database,a comprehensive quality assessment method of fused image based on spectral distortion and spatial distortion is proposed.Firstly,the subjective data in the database is augmented for the comprehensive quality of the fused image.Then,an objective quality assessment method is proposed.The original MS image and the original PAN image are simultaneously downgraded and fused to obtain a fused image.Next,it is divided into two steps to obtain independent spectral distortion index and spatial distortion index.Finally,the spectral distortion index and the spatial distortion index are combined to obtain the comprehensive quality value of the fused image.Testing on the fused image quality database,the experimental results show that the proposed objective assessment method has better performance.