Speckle Interference Fringe Pattern Reduction Technology Research

Electronic speckle interferometry is a kind of whole-field nondestructive testing technology for surface and interior of an object. It is a simple, contactless and whole-field method and has low requirement for the smooth degree of the surface. It has strong anti-interference ability and is widely used in deformation, displacement, and internal loss mesurement, etc. Phase must be extracted from speckle fringe pattern in order to obtain the phase map associated with the phase. However the speckle interference fringe pattern contains high noise, which affects the extraction of phase image. So the denoising of speckle interference fringe pattern is the first and very important step of fringe pattern processing. Effective denoising, structure and gray level maintenance of the fringe pattern is of great importance in speckle interferometry technology. This thesis mainly aims at exploiting the denosing technology for the speckle fringe pattern.The traditional noise reduction method of speckle pattern has simple concept and high computational efficiency, while the noise reduction result is not obvious and it may cause blurring of fringe to a certain degree. Based on the structural characteristics of fringes, spin filtering is proposed and improves the noise reduction to some extent.According to its low efficiency and the advantages of wavelet decomposition, this thesis combined wavelet decomposition with spin filtering to propose an improved spin filtering denoising method, which improved the effect and efficiency of noise reduction by filtering in the low frequency domain of the fringe pattern.When there are solution limits in fixed direction of spin filtering, the structure of fringe pattern is complex, and the fringe density is large, this thesis studies the local frequency domain weighting filtering(Goldstein) algorithm of the speckle interference fringe pattern to improve the spin filtering disadvantage.The algorithm uses the fringe information and the difference of noise in the frequency distribution to highlight the main frequency components of fringe and suppress the image noise. Due to the small amount of noise after the denoising, an improved Goldstein algorithm is proposed to combine with anisotropic diffusion effectively and conduct anisotropic diffusion before the Goldstein, which can reduces some noise and have no effect on the fringe structure. The improved algorithm has great advantages inthe fringe structure and denoising and can effectively solve the accuracy problem during the extraction of the fringe structure in spin filtering.The main innovations of this paper are as follows:1. An improved spin filtering denoising method is proposed, which combines spin filtering with wavelet decomposition and conducts spin filtering in the low frequency domain. On the one hand, the denoising result is better than the traditional filtering method and spin filtering method. On the other hand, the efficiency is obviously superior to spin filtering. The experimental results and the profile show that the improved algorithm is better than the original method in noise reduction result and time. However,due to the defects in the solution direction of the fringe pattern, the fringe direction precision is limited, and when the fringe density is larger, the wavelet transform can lose some detailed information, which may cause some errors when obtaining information related to object variation by further phase extraction of fringe pattern and phase unwrapping.2. An improved Goldstein algorithm is proposed. First the image is processed by anisotropic diffusion filtering. According to different scale area, the image is smoothed in the stripe direction, while smoothing in its edge area is reduced, whi ch can maintain the structure as well as reduce a certain degree of noise. Then Goldstein filtering is used to highlight the stripe frequency components and suppre ss noise by take advantage of the distribution difference of stripe and noise in fre quency band. Experimental results show that the improved algorithm is better than the original Goldstein filtering in both denoising and stripe structure maintenance.Besides, due to its simple principle and easy parameter debugging, the algorithm has a relatively high practical value.