Acceleration of Image Stitching Using Embedded Graphics Processing Unit

Feature detection and matching are powerful techniques used in many computer vision applications such as image registration, tracking, and object detection. In this paper, a parallel implementation for invariant feature point based image warping and stitching using embedded GPU platform is implemented. The proposed solution is a mix of OpenCV functions and Unified Device Architecture (CUDA) kernels. CUDA kernel is used to perform the image translation tasks based on the translation info obtained by OpenCV. A sequential code is developed first to be used as a reference for the speed up calculations. The experimental results show a speed up of 100x and more using our GPU code with large images.