四元数贝塞尔-傅里叶矩及其不变量研究

Moment functions and moment invariants have been widely used in gray-level image analysis and pattern recognition, but these approaches may loose the existing correlation between color components when applying to color image. A new set of moments named orthogonal Quaternion Bessel-Fourier moments (QBFMs) based on quaternion theory was introduced to deal with color image as a vectorial field, the relationship between the new moments and conventional orthogonal Bessel-Fourier moments for the signal channel was given. The invariance to image translation and rotation was investigated and the invariants including phase information were constructed. Experimental results show that the orthogonal quaternion Bessel-Fourier moments perform better than quaternion Zernike moments (QZMs) in terms of image reconstruction capability and noise robustness.