Super-Resolution Reconstruction Of Three-Line-Scanner Images

The Three-Line-Scanner camera can capture three images in the direction of looking forwards,downwards and backwards from the aircraft or satellite in the same time.So we can obtain three images in the same scene.In view of this,we present a maximum a posteriori estimation framework to obtain a high-resolution image from the Three-Line-Scanner images.According to the level of Three-Line-Scanner images,The superresolution process can be devided into two class.One is orthograph superresolution,the other is Level 1 image superresolution.The major work implemented are presented as follows:(1)The mathematics basis of superresolution technique and two model of it are discussed in detail.This paper also discusses the relationship between the superresolution technique and the inverse problems.Moreover,we present the computation method of the superresolution model.(2)This paper introduce the Image Registration mothed excluded aliased frequency domain into the super-resolution reconstruction of Level 1 TLS images.The algorithm can precisely estimate the image registration parameter by excluding aliased frequency domain of the low resolution images and killing the center part of the magnitude spectrum.In the same time,the keren sub-pixel registration method is discussed in detail.In order to overcome out its disadvantage. Moreover,this paper put forward a new improvement approach about keren method and its iterative method.The improvement approach base on the four parameters affine transformation model and abandon the rigid body transformation model.This change avoids the error that is brought on by the tailor series expandedness of angle and improves the precise of image registration.(3)This paper analyse the image capturing procedure of Camera and simulate the gaussian MTF of the static image captured by the Camera.we also introduce error-parameter method and GCV mothod to evaluate the Gaussian MTF of static images.In order to evaluate the MTF of remote sensing images whose MTF do not obey the Gaussian model,this paper introduce the edge method.However, Experimental results show that the error-parameter method and the edge method can not be used to supper-resolution reconstruction model because the MTF which obtained form the low resolution images is different from the MTF of the blured high resolution images. (4)In order to improve the registration accuracy of Level 1 TLS Images,This paper proposes a new optical flow registration method.This approach uses the Normalized Cross-Correlation registration algorithm before using Lucas-Kanade optical flow registration algorithm.(5)This paper present a new super-resolution model to handle the Level 1TLS images.In this model,we introduce the volatile blurs into the gaussian PSF model and take into account the errors of image registration.By Alternating Minimization(AM) algorithm,we can estimate the blur and HR image progressively.Experimental results are shown that our model correspond to the actual TLS images and the AM blind super-resolution approach can be used to enhance the resolution of aerial Images and remote sensing.(6)The MSSIM index to evaluate the result of super-resolution reconstruction is introduced.Based on the TLS images,this paper analyze the reconstruction result of all kind of superresolution algorithm.Experimental results are shown that the robust super-resolution algorithm is better than the other super-resolution algorithms and the MSSIM index is better than the other indexs...