改进块匹配宏块的快速传感器电子稳像方法

Improved Fast Block-matching Algorithm for Sensor Electronic Image Stabilization

提出了一种改进块匹配宏块分布排列的快速传感器电子稳像算法,通过陀螺传感器测量摄像系统的抖动,利用小范围快速块匹配算法估计局部运动矢量,再运用最小二乘法解算全局运动矢量.小范围快速块匹配算法得到的局部运动矢量准确度高,仅需部分局部运动矢量即可准确解算出全局运动矢量.基于此在保证运动矢量准确度情况下,对块匹配宏块的分布排列进行了改进,从而减少匹配宏块数量加快算法速度.通过对宏块网格模型的分析,得出对小范围快速块匹配算法进行宏块分布改进的方案,进而设计出快速传感器电子稳像算法.仿真及实验表明:运算时间提高89%左右,且算法准确度略高于改进前算法.

An improved block density fast algorithm for steadying shaking image was presented.The jitter of camera system was measured by gyro,while the local motion vectors was estimated by the small area fast block-matching algorithm,and the global motion vector was calculated by least square method.Small area fast block-matching algorithm,which has high calculating precision,is good at calculating the local motion vectors.So the least square method can exactly calculate global motion vector by using parts of the local motion vectors.For this reason,the black density can be improved to speed up the algorithm and retain the precision of the global motion vector at the same time.The small area fast block-matching algorithm was optimized by analyzing the block gridding,and then the fast sensor electronic stabilization algorithm was designed.The simulation and experiment show that the algorithm is slightly more accurate and about 89% faster than the small area fast algorithm.