Fast and stable algorithms for high-order Pseudo Zernike moments and image reconstruction

摘要

Pseudo Zernike moments are broadly applied in the fields of image processing and pattern recognition. In this paper, we propose fast and stable methods for calculating high-order Pseudo Zernike moments. A new recurrence is introduced with the addition of a proof. Combining with the Farey sequence, the proposed method is adequate for fast computation. Furthermore, by collaborating with some procedures such as filter method or patch method, we can enhance the accuracy dramatically. The experimental results show that it takes 8.360 s to compute the top 500-order Pseudo Zernike moments of an image with 512 by 512 pixels using the proposed method. Its normalized mean square error is 0.000564363 if 500-order moments are used to reconstruct the image. When computing high-order Pseudo Zernike moments, the proposed filter method surpasses other compared methods in both speed and accuracy.