Research On Image Super-Resolution Reconstruction Algorithm

Image super-resolution technology is a very popular technology,it processes a single or multiple low-resolution degradation image to get a higher quality image.It is extensive in the application of the medical imaging,satellite imaging,video surveillance and other fields.The method based on reconstruction,a class of the technology,is a very important methods.The method requires to establish the mathematical model,we can use some algorithms to solve the problem of fuzzying and denoising,and we need to solve the optimal solution about the objective function across to the model.This paper mainly study the kernel regression method for multi-frame image.The experimental results show the effectiveness and robustness of the proposed method.The research content and innovation of This paper are expressed as follows:This paper summarizes the relevant theoretical knowledge about the image super-resolution reconstruction.This paper introduced several classic algorithms and analyzed their principles and advantages and disadvantages.2.This paper studied Takena's method about single-frame based on steering kernel regression.This paper mainly introduce the principle of classical kernel regression algorithm and kernel regression method.And this paper studied Bilateral kernel regression method and the steering kernel regression method which can improve the short comings of classical algorithm.3.This paper studied the multi-frame image super-resolution reconstruction algorithm.I mainly improved the single-frame steering kernel regression method.I extended this method to multi-frame image super-resolution reconstruction and proposed a multi-frame image super-resolution reconstruction method based on steering kernel regression.Firstly the algorithm is to register the sub-pixels of low-resolution image sequence.Then we use non-uniform interpolation of the steering kernel regression to obtain the initial estimates of the high-resolution images.And we use the iterative steering kernel regression to update the initial estimation and achieve the effect of noise reduction and local structure enhancement.We recover it by using image restoration method based on total variation regularization.Finally,the high resolution reconstruction image is output.4.This paper compared experiment about them in image sequences and real video image sequences,By this algorithm compared with other algorithms,the experimental results show that this algorithm is more robustness,which can keep the detailed information in the image and have a better denoising ability,which can improve the image reconstruction quality.