Research On Termination Method Based On Sensor Electronic Image Stabilization

In recent years, camera system has been being used in many fields. Random jitter cause images obscure, it not only makes observers to feel vertigo, but also affects the image processing, such as target recognition, target tracking, and so on.This paper does much research on early termination method, which is based on sensor electronic image stabilization. The algorithm, including theoretical research, mathematical model, simulation experiments, is introduced. Besides, an experiment system based on personal computer is set up. The study is made actively on both image processing and sensor electronic image stabilization system, and it mainly includes four parts:Firstly, this part does much research on theoretical basis, mathematical models, and experiment system of SEIS. According to the basic work, the problems are proposed, and the research is conducted in next parts:Secondly, a novel early termination method is proposed. It does much work on theoretical research, mainly introduces the mathematical models and advantages. Besides, two simulation experiments, an image covered in small area and an image covered in large area, are made by MATLAB software. The simulation result shows that the method has effect on both situations, the searching step is decreased 18 steps at most, and the accuracy is improved remarkably. The method increases the accuracy of global motion vector information greatly.Thirdly, It uses a gyro to detect the motion situation of camera system and estimate the motion direction. According to the information captured by gyro, the early termination method can find out the singular points of local motion estimation, and then break the operation and get rid of the singular points.Fourthly, an experiment system is set up to prove the method. A gyro is fixed on the camera system, and then set them in a vibration situation. The images captured by camera system are processed on PC. The result exclaims that the operation speed is increased 47.28 per cent; the precision is improved 13.04 per cent.This paper has presented a method which can reject the singular points without acquiring all local motion vector information, so it reduces the operation and increases calculation speed. Moreover, the image stabilization precision has been improved by better consistency of local motion vector set. The work is valuable for low cost, system miniaturization and can be applied to target tracking, stabilizing sighting.