抄録
Moments are widely-used feature extractors due to their superior discriminatory power and geometrical invariance. Unfortunately, moments suffer heavy computational load and result long time spending. In viewing of the problem, we proposed a new technique in using moments- apply moments on wavelet subband. In this study, pseudo Zernike moments are selected as feature extractors due to its enhanced feature representation capability. Implementation of moments on wavelet subband affords advantages of performing local-to-global analysis and decomposing image into lower resolution. Experimental results show that this hybrid achieves computational time reduction of 36.23% with enhanced authentication performance.
