Many practical problems have been solved by deep learning including image super-resolution , video recognition and speech recognition . With the remarkable improvement of computing power, deep learning as a powerful algorithm allows the multi-layer network to learn characteristics of data. Thus, developing an intelligent method to quickly process a large number of fringe patterns with satisfied denoising results, will play an important role in the field of optical measurement based on fringe patterns. Although these existing denoising methods that achieve satisfactory denoising results, these methods will spend much time to process a large number noisy of fringe patterns. Uzanet et al. proposed the nonlocal means filter for speckle denoising of digital holography. Fu and Zhang proposed an image decomposition method to reduce the noise of fringe pattern. Qian , proposed the windowed Fourier filter algorithm (WFF).
Wang et al. proposed a coherence enhancing diffusion method for fringe denoising. However, the noise can cause severe distortion in the estimated phase distribution of the measured object, and denoising is a necessary pre-processing step to analyze the fringe patterns. Many fringe demodulation methods have been proposed to obtain high-precision phase distribution, such as the phase-shifting interferometry and the windowed Fourier transform algorithm . Interferometry as a non-contact high-precision measurement technology can recover the phase distribution from the fringe pattern.
0 kommentar(er)
