CUDA is a type of technology that performs general purposes fast and parallel computation by running tens of thousands of threads concurrently. It makes users develop general GPU programs easily. This paper analyzes the distinct features of CUDA and summarizes the general program mode of CUDA. Furthermore, it proposes and implements an image geometric transform process by the CUDA. By using a position incremental offset method, instead of a large number of multiplications of the original algorithm, and by applying the rapid parallel computing ability of CUDA to transform the geometric image, solve the problem of the inefficient image transform based on CPU. The experimental results show that, as the image size increase, application of CUDA on image geometric transform can reach to more than 100x.