Research On Super-Resolution Reconstruction Of Multi-Frame Images

In recent years,with the development of scientific research and practical technology,the demand for high resolution image and video is increasing.However,in the actual acquisition process,the acquired image and video are usually affected by many degradation factors,which can not meet the needs of practical applications.Super-resolution reconstruction technology,which can improve the spatial resolution of the acquired images and videos by processing them at the existing hardware level,does not need high cost,it can obtain better treatment effect and has high research value,so it has been widely concerned.Multi-frame image super-Resolution reconstruction technology is to reconstruct a High Resolution(HR)image by processing multiple Low Resolution(LR)images with complementary information,it has important application value in security,monitoring,computer vision,military reconnaissance,medical imaging and so on.In the first chapter of this paper,we introduce the significance and status quo of SR reconstruction,and then focus on the two key problems of sub-pixel displacement image acquisition and multi-frame image reconstruction in multi-frame super-resolution reconstruction,the problems of sub-pixel displacement image acquisition,image edge information preservation and multi-frame and multi-frequency reconstruction are mainly studied.On this basis,this paper has achieved some innovative research results,and proposed three multi-frame image reconstruction algorithm:(1)A gradient constrained multi-frame image reconstruction algorithm based on POCS is proposed.This algorithm makes use of the integrity of the edge information of the original low-resolution image,adds the gradient constraint of the edge information in the reconstruction process,and preserves the texture and edge information of the LR image.The gradient constraint method based on edge detection includes three processes:Edge detection,gradient calculation and gradient error calculation.The experimental results show that the Algorithm can reduce the lost of high-frequency information in the process of image reconstruction and make the image clearer.(2)A registration-based multi-frame sub-pixel displacement image reconstruction algorithm is proposed.In this algorithm,SIFT-FLANN Error Rejection(SFER),a registration algorithm which combines SIFT-FLANN with mismatched point Rejection,is firstly used to filter the images,and then the sub-pixel displacement image sequences with known displacement relation are obtained.Experimental results show that the proposed method can obtain sub-pixel displacement image sources with high registration accuracy,and the proposed method can obtain high-resolution images with prominent details and better visual effect,and can correct the "deviation" of the reconstructed image information caused by registration error.At the same time,the effect of reconstruction multiple and low-resolution image on the reconstructed image is analyzed,which further shows the effectiveness of the method.(3)An image multi-resolution reconstruction method based on weighted wavelet transform is proposed.In this method,low-resolution images are decomposed into low-frequency sub-images and high-frequency sub-images by using the multi-resolution property of Wavelet Transform,and multi-frame images are reconstructed by using the image features of high-frequency sub-images and low-frequency sub-images respectively.The experimental results show that this method can mine the feature information of image from different scales by multi-frequency processing,and get the high frequency information of different frequency bands without losing the low frequency information,so as to get the image with clear edge and rich information.