Classification and Recognition of Character Using WP Decomposition, Zernike Moments and Fuzzy Integral

An algorithm was proposed to recognize characters based on optimal wavelet packet (WP), Zernike moments and fuzzy integral. At first, optimal wavelet decomposition of the given character image was gotten by fuzzy rules, and characters in training set were put into n subspaces through optimal wavelet packet. Then, for each subspace, the Zernike moments were employed to get the mean membership grades of test characters with respect to each class in training set, separately. At last, the final class that test characters belonged to was declared according to aggregating the above membership grades and the fuzzy measures using fuzzy integral. The experiment proved that the proposed algorithm had a better recognition effect.