Research On Image Super-Resolution Reconstruction Algorithm Based On Improved POCS

The images acquired in the fields of video surveillance processing,military remote sensing investigation,underwater image processing,and medical image focus inspection are usually vague and have low resolution.However,in actual applications,high-resolution images containing a large amount of detailed information are required.Therefore,improving the resolution of the acquired image or performing image restoration becomes a crucial issue.If the high resolution image is directly obtained by improving the performance of the imaging system,the cost of the imaging device will increase,and if it passes the image super The resolution reconstruction technology indirectly increases its resolution,can obtain better results and the reconstruction process is simple and easy to operate,so it becomes a research hotspot in the field of image super-resolution reconstruction.Image super-resolution reconstruction technology has developed rapidly in recent years.Researchers at home and abroad have proposed a variety of super-resolution reconstruction algorithms,which are mainly divided into two categories: frequency domain method and space domain method.The frequency domain method is aimed at the overall translation between images,so today there is still no major breakthrough in the study of this method.The spatial domain method uses a global observation model that includes global and local motion,optical blur,motion blur,spatial variable point spread functions,and non-ideal sampling.Compared with frequency domain method,spatial domain method has better adaptability and reconstruction effect,so it is a research hotspot in the field of superresolution reconstruction.This paper focuses on Projection Onto Convex Sets(POCS)in the spatial domain algorithm.The POCS algorithm has obvious advantages in intuition,validity,practicality of the observation model,and flexible application of prior information,so the research prospect in the super-resolution reconstruction algorithm is the best.However,the existing POCS reconstruction algorithms still have the following deficiencies:(1)the edges of the image obtained by the bilinear interpolation algorithm are blurred;(2)the correction threshold of the data consistency constraint is a constant,ignoring the object and background noise The difference leads to an edge blurred and low signal to noise ratio image.In this paper,discrete wavelet transform is used to denoise the image preprocessing experiments.The experimental results show that the discrete wavelet transform is better than the salt and pepper noise in the removal of Gaussian noise,but it can only remove the noise without increasing the image resolution,so we need to continue the super-resolution reconstruction of the pre-processed image.In addition,the image obtained by bilinear interpolation using the traditional POCS algorithm will appear edge blurring because of its smooth effect on the image,and a gradient interpolation algorithm is used to perform the initial estimation of the reference frame;due to the correction in the traditional POCS algorithm.The threshold is constant,so that the difference between the object and the background noise is ignored,the same processing mode is used for all the pixels,resulting in the problem of blurred edges and low signal to noise ratio of the image,and the correction threshold is improved to a variable threshold.,you can make different corrections for different pixel areas.The experimental results show that the POCS algorithm based on gradient interpolation and variable threshold improved in this paper reduces the influence of interpolation points on the gray-scale change rate compared to bilinear interpolation,bicubic interpolation,and edge-hold-based POCS algorithm.The information of the edge of the image is maintained,and the purpose of filtering out the noise adaptively is achieved.