Denoising of Nondestructive Examination Data Using Wavelet, Maximum Entropy, and Limited Differential Methods

A simple and original denoising method, the ``limited differential method,'' has been developed. The algorithm is based on iterated local-pixel-averaging, and is very effective for large-amplitude speckled noise on a smoother background signal. For noise of this type, tests on both noisy two- dimensional images and noisy ultrasonic-scattering data volumes clearly demonstrate the superiority of the method relative to three more complicated standard methods: Fourier processing, wavelet denoising, and maximum-entropy reconstruction.