计算变形雅可比-傅里叶矩的改进算法

An improved algorithm of Pseudo-Jacobi-Fourier moments

因其特殊物理性质,不变矩在模式识别中被广泛采用,但多数正交不变矩是在极坐标系中定义的,相关计算常常采取图像函数的极坐标变换,由此带来的量化误差和算法复杂度等尚未得到很好的解决。本文对变形雅可比-傅里叶矩(PJFM′s,Pseudo-Jacobi-Fourier moments)的常用算法进行改进,提出一种直接在笛卡尔坐标系中计算PJFM′s的算法,并将其应用于二值图像的重建实验。实验结果表明,该算法在计算误差和计算速度方面都有明显的改善。

Invariant moments are widely used in the pattern recognition due to their special physics property.But the related computation for orthogonal moments usually adopts the polar coordinate transformation of the image function.Therefore,the problem of quantized error and computation complexity has not been solved well yet.An improved algorithm to compute Pseudo-Jacobi-Fourier moments（PJFM′s） directly in the Cartesian coordinate system is proposed in this paper.Then we apply this improved algorithm to reconstruct the binary image and the experimental results show that the improved algorithm has a better performance in both reconstruction error and calculation time.