一种基于高阶统计量的不变量图像识别算法

Invariant Image Recognition Based on Higher-order Statistics

提出了一种基于高阶统计量的图像识别算法。利用投影将图像数据变换到一维空间,降低了计算的复杂度。通过计算投影数据的高阶统计量,构造出具有比例和平移不变性的图像特征。为得到旋转不变性,提出了一种基于循环相关的算法,从而得到该图像真正具有不变性的特征。还提出了一种改进的整体平均算法,使得该不变量计算速度大为提高。通过仿真对比,指出算法是有效的。

A new image recognition method using higher order statistics is proposed. Projection is used to reduce computation complexity, the scale and shift invariant is constructed in the projection space. To get the rotational invariant, circular correlation is proposed. An improved ensemble average algorithm is studied and the computation speed is greatly improved. Simulation results are given to obtain insight into the efficiencies offered by the method.