Research On Fringe Orientation Estimation And Filtering Methods Of Fringe Pattern Based On Regularized Filtering And Singular Value Decomposition

Optical interferometry technology is widely used in scientific research and engineering practice.The fringe pattern formed in the imaging process includes phase differences before and after the deformation of a measured object,and the measured information of the object is obtained by extracting the phase of the fringe pattern.However,due to factors such as the imaging principle and external environments,latent fringe pattern is contaminated with heavy noise,which has brought difficulties to the extraction of fringe phase.Al-though people have proposed many methods for eliminating noise according to the particularity of fringe pattern,they still cannot meet higher requirement for accuracy in engineering.The fringe pattern filtering and phase extraction are always a hot spot and key issue in the field of international optical mea-surement.Oriented diffusion equation technique carries out a kind of directional filtering of fringe pattern by constructing different diffusions along the tangent and normal directions of the fringe pattern.In this thesis,based on the idea of oriented diffusion equation technique,advanced methods are proposed with more accurate fringe orientation estimation,more robust derivative calculation and more reasonable normal direction diffusion,in the context of different noise of fringe pattern.These methods obtain better results in fringe orientation estimation and fringe pattern filtering,respectively.Specifically,this article has done innovative research works in the following four aspects.1.Regularized singular-value decomposition technique.The gra-dient-based technique is a classical fringe orientation estimation technique;the eigenvector-based technique is a key step in constructing the diffusion tensor in the coherence enhancing diffusion technique;the iterative steering-kernel regression technique uses an adaptive kernel function to denoise image.In this thesis these advanced techniques are synthesized and a regularized singular value decomposition technique is proposed for fringe orientation estimation,which obtains better fringe orientation estimation results.2.Selective second-order oriented diffusion equation technique based on regularized filtering.The second-order oriented diffusion equa-tion technique filters out noise from noisy fringe pattern by directly calculating derivations and estimating fringe orientation.In this thesis a regularized filter is used to reduce the effect of noise on the fringe pattern and to construct a more robust estimation of the first-order and second-order partial derivatives,which obtains a better filtering result.3.Oriented diffusion equation technique based on regularized singular value decomposition.In this thesis,we use the regularized singu-lar value decomposition technique to obtain the estimation of fringe orienta-tion,and we construct a directional diffusion along the more accurate fringe orientation,which obtains a better filtering result.4.Selective second-order oriented diffusion equation technique based on regularized singular value decomposition and normal dif-fusion.In this thesis,a bidirectional diffusion filtering technique is proposed,which uses the selective second-order oriented diffusion equation technique to construct a main diffusion in the fringe orientation estimated by the regular-ized singular value decomposition technique.At the same time,it constructs an auxiliary diffusion in the normal direction using the coherence enhancing diffusion technique,which obtains a better filtering result in fringe areas with different densities.Using advanced techniques proposed in this thesis,a large number of ex-periments are performed on fringe patterns obtained by the electronic speckle interferometry technique,where robust fringe orientation estimations are ob-tained,and noise is filtered out better while preserving the important informa-tion of fringe.These lay a foundation for further phase extraction and accurate measurement,which has important scientific and practical significance.The research in this thesis further strengthens the collaborative innova-tion of computational mathematics and information science,and deepens and enriches the research of fringe orientation estimation and image denoising tech-niques.it is expected to be further extended to other application fields of image processing.