Research On ESPI Image Denoising Technology Based On Improved Total Variation

Images play an important and significance role in the dissemination of information.In recent years,with the continuous improvement of computer technology,the application of image processing in the fields of medicine,optics,agriculture,industry,etc.has been rapidly developed,and the accuracy of image processing is becoming higher and higher.In the field of electronic speckle pattern interferometry(ESPI),reducing the noise of images is of great significance.ESPI is widely used in full-field non-destructive testing inside and on the surfaces of objects due to its simplicity,contact-less full-field real-time measurement,and strong anti-interference ability.Real-time detection of deformation,internal loss and displacement.The acquisition of the speckle phase map requires phase extraction from the ESPI image,and the presence of a large amount of intrinsic noise affects the accurate phase extraction.Therefore,the filtering of the speckle fringe pattern becomes the primary part in the speckle fringe pattern processing.The integrity of the fringe pattern and the image edge information should be maintained while filtering noise as much as possible.This thesis mainly studies the noise reduction technology of ESPI images.Traditional ESPI image denoising method is simple and easy to operate,but for the complex noise in the speckle fringe pattern,the noise filtering effect is not obvious.Not only is there more residual noise,but also the denoising image is blurry.The TV(Total Variation)denoising model with anisotropic diffusion characteristics has a good noise filtering effect,but the staircase effect occurs in the process of noise reduction.In this thesis,an improved TV denoising method is proposed by making advantage of TV denoising model and wavelet decomposition.Wavelet decomposition is used to reduce some noise in the image,which can improve the image denoising effect to some extent and the noise reduction efficiency.At the same time,this thesis also studies the adaptive total variation denoising model,which can adjust the parameters according to the noise level adaptively and better balance the regularization term and the fidelity term.In order to make the image information more complete and reduce the staircase effect of the adaptive total variation noise reduction model,this thesis proposes an adaptive total variationfiltering method based on BM3D(Block-Matching and 3-D filtering),which can protect the edge information of the image more effectively and solve the problem of image information loss caused by the improved TV denoising method.The BM3D-based adaptive total variation denoising model can maintain the integrity of the fringe structure and the edges features of the fringe pattern while eliminating the noise.For the fringe pattern with more severe noise,a better noise reduction effect can also be obtained.The main innovations of this thesis are:1.An improved TV denoising method is proposed,which combines The TV denoising model and wavelet decomposition to make full use of the low frequency fringe pattern.Simulation experiment results show that the improved method has better noise reduction performance than the traditional filtering method,and the efficiency is also obviously improved.However,some information of the image is lost after wavelet decomposition,which may lead to the information incompleteness of the image itself.This defect will affect the phase extraction and unwrapping of the future fringe image.2.An improved adaptive total variation denoising method based on BM3 D is proposed.After BM3 D smoothing,the electronic speckle interference fringe pattern is filtered with the adaptive total variation denoising model by taking the processed image as a new fidelity term.Experiment results show that the proposed method has better noise filtering effect,and the structure of the speckle fringe pattern is well maintained,which can reduce the appearance of the staircase effect and can preserve the details,texture and edge information of the image at a large level.