Interferometric techniques are very important in the metrology field, while the quality of the interferogram will directly affect the retrieval phase of the tested object. This paper presents a method to improve the quality of the interferogram including restoration of noise aliasing and moiré distortion by using the Gaussian Process Regression (GPR). Through choosing a suitable covariance function to describe the relationship between points and points in the fringe pattern, we build a Gaussian process regression model of interferogram, denoise the interferogram and improve the resolution at the same time. The treated interferogram can predict and compensate the part of the fringe distortion and enlarge the depth range of the interferometric measurement. Besides, with the resolution elevated of the hologram, a wider spectrum range can be obtained. In order to verify the possibility of this method, several simulations have been done, which showed a good performance in the enhancement of the quality of interferogram.
Fringe pattern denoising is an important process for fringe pattern analysis. In this paper, fringe pattern denoising using the convolutional neural network (CNN) is introduced. We use Gaussian functions to generate the various phase distributions, and then the required training samples are simulated according to theoretical formulas. The noisy fringe pattern can directly obtain the clean fringe pattern using the trained model. The denoising performance has been verified, which can recover high-quality fringe pattern.