Real-time Optical Image Stabilization using High Frame Rate Video Sequence at 500 Fps

摘要

In this paper, we have developed Real-Time optical image stabilization system which uses high frame rate (HFR) video sequence captured by high speed camera head in order to stabilize jitter due to undesired camera motion while handling camera head by handheld system or mounted on mobile robots or vehicles. Optical Image Stabilization (OIS) is conducted using 512×512 color images captured at 500 fps by generating counter motion against high displacement motion. An improved feature extraction algorithm is implemented on high speed vision platform which is facilitated with Field Programmable Gate Array (FPGA) in order to implement user defined hardware logic to reduce the computation cost and to realize real-time image processing. Extracted features are matched with consecutive frames to obtain frame by frame displacement and direction of motion. Based on frame displacement camera fluctuation is calculated. Counter motion is determined by filtering high displacement using Infinite Impulse Response (IIR) high pass filter. High speed camera head is mounted on 2-DOF Pan-Tilt-compensation system which is used to respond to the calculated counter motion. Since the whole process is realized in real-time, high speed OIS is proved as computationally efficient.

出版物
In Proceedings of the 2015 Conference on Advances In Robotics