Cuda Technology And Its Application In Digital Image Mosaic

CUDA technology is a parallel computing architecture developed by NVIDIA, its characteristic is to use the GPU to implement multiple threads concurrently to achieve large-scale, high-speed, general calculation. Its simple development process and programming style, efficient multithreaded parallel programming model can make the computation intensive task take better use of GPU powerful parallel computing resources effectively. And digital image mosaicing is a kind of important compute-intensive applications, the task of it is to map with overlapping area in the images and mosaic into a picture of a higher resolution image. The main steps and key technology including image preprocessing, registration and fusion, etc. The technology have involved all need to make further research and improvement.This paper mainly studies about the CUDA technology and its application in digital image mosaic. Here we have described the principle of multithreaded programming and the method of development in the use of CUDA, and have discussed the basic principle of image mosaic and its key technology.We have selected the SIFT feature matching algorithm to realize the characteristics of image matching in the comparative and analysis of the current image matching algorithm, multi-resolution image fusion algorithm have been selected to realize the image fusion; At last, combined with the above two kinds of algorithm and programmed to realize high speed digital image splicing in CUDA architecture.The main features of this paper are to use the robust SIFT feature matching algorithm based on improved algorithm for feature points extraction and matching and using multiresolution fusion algorithm to realize the seamless image mosaic. In this paper, the above algorithm are all realized in the use of CUDA parallel computing technique and traditional single CPU serial program at the same time, and verify the CUDA programming to enhance the efficiency of image mosaicing through the comparison of time in two implementation scheme. The experimental results show that the proposed algorithm and the improved SIFT algorithm have been realized in the use of CUDA is a good way to solve the matching problem in the process of image mosaicing, improved the efficiency of image matching, provided high quality fusion material for the subsequent image fusion. The results of this paper reflect the CUDA technology is productive in digital image processing of the practical application,has a broad prospect in application.