电子稳像算法研究与实现

【作者】 杨毅；

【导师】 张毅；

【作者基本信息】 南京理工大学 ， 光学工程， 2009， 硕士

【摘要】 电子稳像在大规模集成电路高速发展的今天应用越来越广泛,有效地弥补了光学稳像和机电结合稳像方法的不足,而作为电子稳像基础的算法研究则显得尤为重要,对于各种场合的目标跟踪或侦察是非常具有实际意义的。本论文在分析稳像技术以及常见的几种稳像方式的基础上,研究了电子稳像技术的基本原理,电子稳像的数学模型以及电子稳像的主要特点和主要步骤;分析了几类常见的电子稳像算法,并在对各种算法深入研究的基础上,基于灰度投影算法的原理,利用软件仿真了稳像系统,其中主要包括图像预处理、运动矢量估计、运动矢量平滑以及运动补偿等几个部分;在对现有算法进行研究和仿真的基础上,创新性地将顺向补偿和反向补偿的运动补偿方法互相融合,并对原有补偿方法进行了改进,防止在稳像结果中出现虚假的静止场景;创新性地采用最小视域的方法,减少了由于补偿造成的黑色边框造成的不稳定感。仿真结果表明实验中所采用的灰度投影算法在实时性和精度方面均可以满足要求。论文最后对算法和仿真过程中出现的问题进行了细致的分析,对课题的下一步研究方向提出了建议。更多还原

【Abstract】 Nowadays with the rapid development of VLSI, Electronic Image Stabilization (EIS) has got more widely applied and made up for Optical Image Stabilization and Mechanical Image Stabilization’s shortages effectively. So the algorithm research considered as the foundation of EIS is extraordinary important for the object detection and target tracking in many occasions.In this paper, on the basis of analyzing Image Stabilization Techniques and a few familiar Image Stabilization Modes, the theory and the mathematical model of Electronic Image Stabilization (EIS) are studied, as well as main characteristics and steps of EIS. Several normal algorithms are introduced and EIS System is emulated in MATLAB software environment based on the basic principle of Gray Projection Algorithm and intensive study of several kinds of algorithm, including Image Pre-processing, motion vector estimation, motion vector smoothing and motion vector compensation. During algorithm emulating, two basic motion vector compensation methods are combined creatively avoiding fake fixed scene in the result of stabilization. Minimal visual field is creatively used in experiment to reduce instability caused by black compensation region. The Gray Projection Algorithm adopted in this experiment has been proved to meet the demand of accuracy and real time character. In the end of the paper, the problems appear in experiment are analyzed deeply and further research direction is suggested.更多还原