Ultra Complex Integration Of Remote Sensing Images And Its Evaluation Method

Along with the development of remote sensing technology, multispectral remote sensing image as an important type of remote sensing data is formed. As the key point of multispectral remote sensing image fusion study, multispectral and panchromatic images fusion, anticipants a fusion result with high spectral resolution and high spatial resolution. To achieve this goal, hypercomplex is introduced.First of all, existing image fusion approaches oftenly lead to color distortion of fused images. To resolve this issue, hypercomplex matrices are introduced to describe panchromatic image and the multispectral images respectively. Every Hypercomplex vector pixel binds specified geometric relationship of components of multispectral image. It is treated as an element of hypercomplex algebra. The hypercomplex singular value decomposition approach is applied to the hypercomplex matrices of panchromatic and multiple spectral images, then hypercompelx singular values representing the feature information of two matrices are obtained. A hypercomplex principle component weighted approach to multispectral and panchromatic images fusions is presented in this dissertation to resolve color distortion of panchromatic image and the multispectral images fusion.Analytical results show that the residual error of high resolution multiple spectral image can be recovered, when the simplex part via applying hypercomplex symplectic decomposition to the residual error of low resolution multispectral images is replaced by the residual error of high resolution panchromatic image. In this way, multispectral and panchromatic images fusion can be implemented by synthesis of symplectic decomposition. In this dissertation, a residual error hypercomplex symplectic decomposition approach to multispectral and panchromatic images fusion is presented. Experimental results demonstrate simplicity and efficiency of proposed method.Secondly, to achieve better evaluation of fusion image quality, objective methods of color image quality assessment are studied, and applied to evaluate quality of remote sensing image fusion. When a color image is degraded by image compression, noise, transmission error, and so on, the hypercomplex gradient information of image is changed accordingly. And the hypercomplex Harris response is also changed. Hence, the degree of change in hypercomplex Harris response is related to quality degradation of color image. Hypercomplex gradient information matrix is presented to compute Hypercomplex Harris response of color image. Then in this dissertation, a hypercomplex Harris Response-based Image Metric (HHRIM) to evaluate color image quality is proposed.In order to comprehensively evaluate the distortion types from the structural or color information, a hypercomplex Edge Structural Similarity color image quality index is then presented. The proposed index is designed by modeling any image distortion as a combination of factors:loss of correlation, luminance distortion, contrast distortion, color distortion. The experimental results of various color image distortion types indicate that the proposed index not only performs significantly better than the known indexes on a grayscale image, but also measure whether the distortion types are from the structural or color information of a color image.Finally, apply above evaluation methods to fusion results of the IHS, PCA, wavelet-based methods and the presented approaches above, and the results are conformed to the existing evaluation parameters. Moreover, these methods have achieved simplicity and high reliablity. The evaluation result also demonstrate the efficiency the proposed fusion methods.