Fast Image Matching Based on GPU Parallel Computing
Abstract：Based on the speedup theories of traditional method of image matching algorithm using CPU and GPU,a more effective GPU parallel computing algorithm is proposed.The algorithm introduces a new design on blocks and threads to implement the parallel architecture of CUDA and takes full advantage of high-speed shared memory to get maximum performance.GPU can execute faster by minimizing the access to low-speed memory.The experiment shows that on the basis of current common devices,the processing speed of the optimized algorithm is 48 times and 7～12 times faster than that of the CPU-based implementation and former GPU-based implementation respectively.Compared with the former GPU-based implementation,the improved algorithm also increases the adaptability when dealing with lager image data and using multi-GPU programming.
Keywords：image matchingGPUCUDAparallel computingshared memory
Publication Date：2011-01-01
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：5( 306-310 )