Research On Kalman Filter-based Digital Image Stabilization Algorithm

Digital image stabilization technology is a hot issue of image stabilization technology. Based on the in-depth research of the key technologies of digital image stabilization, pointed out the excessive calculation problem of global motion vector estimation using block matching method, and the image of the processed video often presence large undefined region when using the classical Kalman filter-based digital image stabilization algorithm to remove the jitter in the video which exist complex intentional scanning motion. To solve this problem, we presented an improved digital image stabilization algorithm. Using the block matching method to estimate the global motion vector between two adjacent frames, reduce computation by using Kalman prediction and cross-correlation theorem; filtering the global motion vector with Kalman filter, construct a deviation which can characterize the size of the undefined area, adding it to the residual of the filter as the new residual, judging whether the camera intentional scanning state is changed by the the size and change of deviation, adjusting the value of new residual with fuzzy control method. Experimental results show that this method can effectively filter out the high and low frequency jitter noise in the video, while significantly reducing the undefined region of the image in the processed video.