基于特征点跟踪的数字稳像算法  

Digital image stabilization algorithm based on feature points tracking

作　　者：刘栋[1] 赵跃进[1] 尹德森[1] 郭亮[1] 武红[1] 刘小华[1] 

机构地区：[1]北京理工大学信息科学技术学院光电工程系,北京100081

出　　处：《光学技术》2008年第2期281-284,共4页Optical Technique

摘　　要：针对图像平移和旋转的数字电子稳像问题,提出了一种新的稳像算法。采用K-R法(Kitchen-Rosenfeld)提取图像特征,根据Similarity变换模型进行运动估计,将图像平移和旋转模型转化为仅绕中心点旋转的单一旋转模型,降低了模型的复杂度和计算量。实验证明,这种算法具有很好的稳像效果。A new algorithm is presented to solve the image rotation and translation of the digital image stabilization（DIS）system.The algorithm uses K-R method to acquire feature of the image,Similarity transform model is used in motion estimation.The model that has rotation and translation becomes another one which is only a rotation model rotating around the center.The complicacy and calculation of the model are reduced.The result shows that it has favorable stably effect.

关 键 词：K-R法 运动估计 数字稳像 

分 类 号：TH706[机械工程—仪器科学与技术]