| Electronic image stabilization, as an economic and effective image stabilizing approach, has received much attention during these years. A thorough research is made on electronic image stabilizing algorithms and a novel method is proposed that can deal with inter-frame image stabilization.First, this paper made a literature on the development of image stabilizing techniques at home and aboard, and also gave a brief introduction to all the steps required in electronic image stabilization including corner detection, description, matching of corner features, elimination of false matching and correction.Subsequently, methods and algorisms adopted in this paper are discussed. FAST (Features from Accelerated Segment Test) algorithm is used to increase the speed of detecting the corners, and the experiments shows FAST is indeed faster than conventional corner detection algorithms. An improvement is further made in the paper on FAST algorithm to enhance its stability of detection.3×3 SURF description is combined with LBP in the description of corner feature given that LBP can accelerate matching speed of corners. The false matches of feature points are removed by RANSAC to ensure a high matching accuracy, and the frames witch to be corrected are transformed by perspective transformation.Experiments are carried out to test the speed and feasibility of the proposed algorithms. The image after stabilization is compared with reference frame, and PSNR analysis is made to validate the effectiveness of the algorithm.Then, a possible way is put forward that can further improve image stabilizing algorithm mentioned in the paper. The idea is that the number of feature points extracted can be varied according to the characteristic of image, and information content can be increased by merging reference frame and the current frame after stabilization.All the algorithms are compiled with C language in VS2008 C++, making them suitable for being migrated to different platforms. |
PDF Full Text Request