C++ OpenCV Motion Tracking Ping Pong Application GPU accelerated on Nvidia Tegra Board Linux

We are writing an OpenCV C++ application that motion detects a ping pong ball and tries to predict the motion of the ball as it moves in front of the camera to determine if it will go over the net or not.

Our functions utilize the GPU cores to do the image processing. Our problem is that we can't display the OpenCV matrix using imshow since it is painfully slow. Are there any alternatives or suggestions you would have for displaying an OpenCV matrix on Linux, specifically the Nvidia Jetson TK1? OpenCV for Tegra doesn't support OpenGL imshow unfortunately and we can't get the standard OpenCV to compile on the Tegra board with OpenGL.