Digit recognition using trispectral features

Features derived from the trispectra of DFT magnitude slices are used for multi-font digit recognition. These features are insensitive to translation, rotation, or scaling of the input. They are also robust to noise. Classication accuracy tests were conducted on a common data base of 256256 pixel bilevel images of digits in 9 fonts. Randomly rotated and translated noisy versions were used for training and testing. The results indicate that the trispectral features are better than moment i n v ariants and ane moment invariants. They achieve a classication accuracy of 95% compared to about 81% for Hu's moment i n v ariants and 39% for Flusser/Suk ane moment i n v ariants on the same data in the presence of 1% impulse noise using a 1-NN clas-sier. A m ultilayer perceptron with no normalization for rotations and translations yields 34% accuracy on 1616 pixel low-pass ltered and decimated versions of the same data.