A real-time stereo rectification of high definition image stream using GPU

Stereo vision systems are essential for the 3D object reconstruction, industrial automation, telerobotics and many more. Stereo rectification process aligns the left and the right images such that their respective image rows are accurately aligned with each other. In this paper, we propose and implement a real-time stereo rectification of the high definition stereo image stream coming from the IP cameras for the comfortable stereoscopic perception. We utilize the GStreamer multimedia framework for the image acquisition. We make use of the GPU for the real-time stereo rectification using the OpenGL. We make use of the OpenGL concepts like the framebuffer object, the vertex buffer object for the real-time image remapping. We divide stereo rectification process into two phases. Initially we compute the stereo rectification inverse lookup maps once using the OpenCV library. Subsequently we apply these inverse maps to each incoming image frame using the OpenGL concepts. We compare our results with the OpenCV realization using the optimized AVX instructions and the CUDA framework. Our OpenGL based module performs the stereo rectification of the full high definition stereo image stream at about 70 frames per second.