CUDA Accelerated Real-time Digital Image Stabilization in a Video Stream

David Pacura, Martin Drahansky
2016 International Journal of Software Engineering and Its Applications  
The most important step for successful video processing in computer vision is its stabilization. Often, it is required to process high resolution video in a real-time. In this paper, a new method for a real-time digital image stabilization in a video stream is presented. This method preserves the intended camera motion and exploits computing power of GPGPU by utilizing CUDA programing interface. In order to reduce required computation power, local search windows are used for the correspondence
more » ... earch of consecutive video frames. These windows are further processed using Local Binary Patterns, which enables fast correlation using bitwise XOR. The experiments on video sequences from both car-mounted and hand-held camera have demonstrated the effectiveness of this method. The speed of stabilization designates this method for video preprocessing in real-time applications. "walking" This video sequence was created by hand-held compact camera during walking through hall with bad luminance conditions. Therefore, it contains blurry frames and changing parallaxes. "car-ride" This video sequence was recorded from camera mounted inside car. Shooting was performed during sunny weather. However, video contains focus changes between infinity and interior and rapid shaking caused by car going through potholes. "object-tracking" This video sequence is shot with hand-held camera with F=500 mm. It tracks movement of distant object during good light conditions. "pan&zoom" This video sequence is also shot with hand-held camera. However, this time with the "panorama" effect (smooth position changes) and big zoom changes.
doi:10.14257/ijseia.2016.10.6.10 fatcat:rtqn374jhvh7dh6ydebpqbfyq4