Implementation Of A Parallel Algorithm Of Remote Sensing Image Fusion Based On Variational Method

With the rapid development of satellite remote sensing technology,remote sensing technology has played a huge role in the national economy and military fields.Pixel-level fusion of remote sensing images(also known as pansharpening)has always been a hot issue in the research of remote sensing technology.It is a technology that fuses high-resolution panchromatic images with abundant spatial information with low-resolution multispectral images to generate high-resolution multispectral images.The traditional remote sensing image fusion algorithm cannot balance the retention of spectral information and spatial information.In recent years,remote sensing image fusion algorithm based on variational method has gradually developed into a new important branch in the field of remote sensing image fusion because of its excellent fusion effect and high flexibility.But at the same time,the variational image fusion algorithm has high computational complexity and low computational efficiency,and can’t cope with the increasing expansion of remote sensing image data.At the same time,with the emergence of high-performance GPU and CUDA,the use of GPU for parallel computing has become an increasingly popular means in remote sensing image processing.In order to solve the problem of low operational efficiency of variational remote sensing image fusion algorithm,based on the in-depth study of nonlocal relational variational model,the CUDA is used to process remote sensing image fusion algorithm on GPU.In the parallel algorithm designed in this paper,the multispectral image is pre-processed by interpolation.Then IHS algorithm is used to fuse multispectral image and panchromatic image as the initial image of nonlocal variational image fusion algorithm.In the nonlocal variational image fusion module,three constraints and iteration formulas of the nonlocal variational model are designed in parallel.In the above steps,texture memory,constant memory,shared memory and other optimization methods at different levels are used synthetically.In this paper,multiple sets of remote sensing images with different bands and different sizes are used to evaluate the nonlocal variational remote sensing image fusion parallel algorithm from the aspects of fusion quality and computational efficiency.In terms of fusion quality,this algorithm performs better than other common remote sensing image fusion algorithms in subjective vision and objective indicators.In terms of operational efficiency,the parallel algorithm has significantly improved the computational speed compared with the serial algorithm,and obtained a good speedup ratio.