Ultra-Fast Zernike Moments using FFT and GPU
This work addresses a domain-specific problem for machine vision applications by providing a faster and more stable method for computing Zernike moments, though it is incremental as it builds on existing FFT and GPU techniques.
The paper tackles the slow implementation and numerical stability problems of Zernike moments in machine vision by proposing a novel method using Fast Fourier Transform (FFT) and GPU computing, achieving real-time computation for 4K resolution images and improved numerical stability compared to other methods.
Zernike moments can be used to generate invariant features that are applied in various machine vision applications. They, however, suffer from slow implementation and numerical stability problems. We propose a novel method for computing Zernike using Fast Fourier Transform (FFT) and GPU computing. The method can be used to generate accurate moments up to high orders, and can compute Zernike moments of 4K resolution images in real-time. Numerical accuracies of Zernike moments computed with the proposed FFT approach have been analyzed using the orthogonality property and the results show that they beat other methods in numerical stability. The proposed method is simple and fast and can make use of the huge GPU-FFT libraries that are available in several programming frameworks.