A new effective image super resolution (SR) algorithm which is a hybrid of multiple frame Variational Bayesian (VB) reconstruction and single frame Dictionary Learning (DL) reconstruction method is developed to reconstruct a high resolution (HR) satellite image in this article. Firstly, by employing a variational Bayesian analysis, the unknown high resolution image, the acquisition process, the motion parameters and the unknown model parameters are built together in a single mathematical model with a Bayesian formula, and then the distributions of all unknowns are jointly estimated. Without any parameter adjustment, an HR image is adaptively reconstructed from multiple frame low resolution (LR) images. Secondly, by taking the above HR image as input, a higher resolution image can be rebuilt utilizing the statistical correlation between the HR and LR images which is obtained via the DL method. The VB method effectively uses non-redundant information between LR images to recover HR satellite images. Benefit from the dictionary training of magnanimity image, the DL algorithm is able to provide more high-frequency image details, which means this hybrid of VB and DL method combines the above advantages. The experiments show that this proposed algorithm can effectively increase the image resolution of remote sensing images by 0.5times at least comparing with low resolution image.
Aiming to realize super resolution(SR) to single image and video reconstruction, a super resolution camera model is proposed for the problem that the resolution of the images obtained by traditional cameras behave comparatively low. To achieve this function we put a certain driving device such as piezoelectric ceramics in the camera. By controlling the driving device, a set of continuous low resolution(LR) images can be obtained and stored instantaneity, which reflect the randomness of the displacements and the real-time performance of the storage very well. The low resolution image sequences have different redundant information and some particular priori information, thus it is possible to restore super resolution image factually and effectively. The sample method is used to derive the reconstruction principle of super resolution, which analyzes the possible improvement degree of the resolution in theory. The super resolution algorithm based on learning is used to reconstruct single image and the variational Bayesian algorithm is simulated to reconstruct the low resolution images with random displacements, which models the unknown high resolution image, motion parameters and unknown model parameters in one hierarchical Bayesian framework. Utilizing sub-pixel registration method, a super resolution image of the scene can be reconstructed. The results of 16 images reconstruction show that this camera model can increase the image resolution to 2 times, obtaining images with higher resolution in currently available hardware levels.
The low resolved satellite images caused by serious degradation in remote sensing weaken its utilities in practice. An effective algorithm of high resolution remote sensing image reconstruction is proposed to recover the degraded images using a precise estimated modulated transfer function (MTF) of the imaging system from a curve knife edge. The curve edge is chosen automatically and robustly among many candidate edges, which can provide a higher precision in comparison to straight edge. To suppress the artifacts and noise, the total variation (TV) method is applied as well. The experiments show this algorithm is suitable to recover a high-resolved image with a high signal-to-noise ratio (SNR).