采用GPU并行架构的基于互信息和粒子群算法的异源图像配准

摘要

For visible and infrared images registration, in order to get results gradually by changing resolution from low to high, mutual information is used as similarity measure condition, and the particle swarm optimization algorithm is used to search for the maxima in searching space. Experiments show that the algorithm improves the computational efficiency without reducing the accuracy. For the parallelism in particle swarm optimization algorithm, mutual information algorithm, and affine transformation, a heterogeneous GPU-CPU programming is achieved with CUDA C. Experiments show that the algorithm improves the computational efficiency without reducing the accuracy and affects the extreme values of speedup ratio. It proves that the algorithm has a high correct matching rate, high robustness, and high computational efficiency.%对于异源可见光与红外图像配准,以互信息为相似性度量条件,以粒子群算法为搜索算法,在搜索空间中搜索极大值点,通过改变图像分辨率,由粗到精,逐步实现可见光图像与红外图像的配准,对粒子群算法、统计互信息和仿射变换3个部分的内在并行性,利用CUDA C语言实现GPU-CPU异构编程.实验证明,在不降低精度的前提下提高了算法效率,得到了很好的加速比,算法正确匹配率高,鲁棒性好,计算效率高.