Electronic Image Stabilization Based On Persistence Of Vision Effect

The camera systems are more and more widely used in military or civilian.Interference generated by the unconscious shaking of the camera platform led to the imagesequence jitter, affecting further image analysis. That is why image stabilization start up.To compare with the traditional stabilization methods, the electronic image stabilizationhas much advantage, such as high image stabilizing precision, compact size, light weight,low power consumption and reasonable price.This thesis at first describes the significance of electronic image stabilization and thedevelopment of it in detail. Then it describes the basic principles of electronic imagestabilization, system composition, and image motion model. After that, It analyses theapplication and limitations of the basic algorithms for the electronic image stabilizationwhich includes motion estimation, motion decision, motion compensation.The current methods of electronic image stabilization are sufficient for mostapplications, but they don’t work in the case of image sequences with large jitter. In thiscase, we propose a new way to handle the problem above, which is the electronic imagestabilization method based on the persistence of vision effect. The method uses a mosaicof large scene as a reference frame. And then starts motion estimation in two waysincluding an improved block-matching and feature matching. After that we removeundesired motion with kalman filter. At last, we flash the reference frame in a ImageFusion Algorithm based on the Relative Activity in Pixel Neighborhood. Until now, wecan get a clear and stable frame from the reference frame.In the end, experiments were carried out with a variety of video. Experiment showthat the presented method can access to high-precision image stabilizing effect in real time.Also, this method is effective with scale, rotation, and shift.