作者: Xuling Luo , G. Mirchandani
DOI: 10.1109/ICASSP.2000.862058
关键词:
摘要: This paper presents a novel method for classifying an image into one of predefined classes in data bank by applying mutual information to representation the Fourier amplitude domain. Template and test images are made translation rotation invariant through Fourier-Mellin transform. While could be employed here, we choose instead apply it lower dimension phase spectrum generated complex multiresolution wreath product transform spectrum. The this adequately preserves edges even at resolutions while permitting same time, reduction computational burden. Brodatz textures ORL faces used demonstrate capability algorithm.