Zernike moments, as a representative orthogonal moment, have been widely applied in the fields of image processing and pattern recognition. The calculations are time-consuming due to the complexity of definition. Based on the GPU octant symmetry algorithm in our previous work, this paper presents a novel algorithm to increase the resource utilization by the combined kernel. Also, it optimizes radial polynomials of Zernike moments to reduce amount of calculations. The experimental results demonstrated that the proposed algorithm achieved overall computational performance improvement for any sized images. Moreover, there is no compromise in terms of precision compared to the typical accurate algorithm.

作为代表正交矩的泽尼克矩已被广泛应用于图像处理和模式识别领域。由于定义的复杂性，计算非常耗时。基于我们先前工作中的GPU八分对称性算法，本文提出了一种通过组合内核提高资源利用率的新算法。而且，它优化了Zernike矩的径向多项式，以减少计算量。实验结果表明，该算法对任何尺寸的图像都实现了整体计算性能的提高。而且，与典型的精确算法相比，在精度方面没有任何妥协。