Theoretical Analysis Of Super-resolution Sub-image Stitching Based On Lagrange Multiplier Metho

Super-resolution image reconstruction aims to reconstruct a high-resolution image with clear and detailed features from a given low-resolution image.It is a key research task in the field of computer vision and image processing.Extremely extensive practical value[1].However,at present,researchers have found that the difficulty of super-resolution reconstruction in different regions of low-resolution images is different.Using a unified super-resolution network(large-scale/small-scale)to process low-resolution images will either waste computing resources and hardware,resources(for large scale),or it is difficult to obtain the desired superresolution reconstruction effect(for small scale).To this end,researchers started to process different image regions with different processing strategies.Xiangtaokong et al.cut the lowresolution image into sub-images according to the difficulty of super-resolution reconstruction,and then used a large-scale super-resolution network to process the sub-images that are difficult to super-resolution reconstruction,and used a medium-scale super-resolution network to process super-resolution Reconstruct the sub-images with moderate difficulty,use a small-scale superresolution network to process the sub-images with low super-resolution reconstruction difficulty,and finally stitch these processed sub-images into a complete super-resolution image.The above research methods mainly focus on the cropping of different regions in low-resolution images and the super-resolution image reconstruction of low-resolution images in different regions,and how to seamlessly fuse the super-resolution images reconstructed from low-resolution images in different regions into one The complete super-resolved image is just taken by simple weighted average fusion.Although this fusion method can also obtain satisfactory results,it is mostly set through multiple experiments,the generalization is not strong,and the selection of weights also lacks theoretical guidance.For this reason,based on the weighted average image fusion algorithm,this paper proposes a Lagrangian objective function for super-resolution sub-image fusion,and optimizes the objective function with the help of Lagrange multiplier method to provide super-resolution sub-image mosaic The theoretical reference value of the weight provides a theoretical reference for the weight setting of super-resolution sub-image stitching and fusion,and also improves the universality of the super-resolution image reconstruction algorithm.The main work of this paper includes the following two aspects.First,for the problem of image-level image fusion,we propose a Lagrangian function for image-level super-resolution sub-image fusion and optimize the function by the method of Lagrangian multipliers.The research results provide a theoretical reference for the weight setting of super-resolution sub-image stitching.Second,for the problem of pixel-level image fusion,this paper proposes a Lagrangian function for pixel-pixel super-resolution sub-image fusion and optimizes the function through the method of Lagrangian multipliers.The research results provide a theoretical reference for the weight setting of super-resolution sub-image stitching.