Directional filters for fringe pattern denoising

Abstract

For a successful phase demodulation it is important to have a good quality fringe pattern image. For this reason preprocessing fringe patterns is, many times, an unavoidable task. Often, noise removal is the main problem to be solved, however, the use of ordinary linear filters is not always a proper procedure specially in the presence of high density fringes because the signal and noise are mixed in the Fourier space. Also, as fringe pattern images are two-dimensional functions, frequencies are two-component vectors which requires consider the filtering direction. We present a new denoising technique for preprocessing fringe pattern images which requires to previously estimate the fringe orientation. For cases of high noise levels we modify the proposed technique by means of a regularized local cost function in order to get a better noise response. We present a noise response analysis of the proposed technique, some experimental results and its application to wrapped phase maps denoising.