Image recognition with the discrete rectangular-wave transform

The discrete rectangular-wave transform (DRWT) is obtained by replacing the sines and cosines in the discrete Fourier transform (DFT) with the bipolar rectangular-wave function. The DRWT matrix contains no numbers other than ±1, ±j, and their combinations. There are also both parallel and sequential fast algorithms and architectures for its electro-optical or electronic implementation, which are much faster and easier than fast-Fourier-transform algorithms and architectures. In recognition tasks, the DRWT is used mainly as a feature extractor and is followed by a classification classification algorithm. In experiments involving shape and image recognition, the DRWT is observed to have better performance than the DFT, especially when the classes of objects to be recognized are similar. The DRWT can also be used in a number of ways for distortion-free recognition.