A noise-robust algorithm for classifying cyclic and dihedral symmetric images

A noise-robust algorithm for detection and classification of cyclic and dihedral symmetric images is presented in this paper. For a symmetric image corrupted by an additive white Gaussian noise (AWGN), the proposed algorithm is implemented by converting the symmetry information into the representation of angularly evenly spaced zero-crossing lines in Mexican-hat wavelet domain; in addition, a continuous Mexican-hat ridgelet is applied to detect those zero-crossing lines, which achieves a simple and fast discrimination between cyclic and dihedral symmetries. Experimental results show that the proposed algorithm is very robust against noise and it can automatically classify the cyclic and dihedral symmetric images.