Research On Oriented Diffusion And Sparse Filtering For Fringe Patterns Enhancement

With the rapid development of information society, Electronic Speckle Pattern Interferometry(ESPI)and Interferometric Synthetic Aperture Radar technology(InSAR) are the new interferometric techniques today. ESPI could be used in overall optical measuring for small deformation and morphological. It has non-contact, strong anti-interference and high sensitivity advantages. InSAR use synthetic aperture radar images interfere, it could be used in three-dimension terrain and topography elevation information which is carry on all-weather and wide range of high-precision in remote sensing measure. ESPI belongs to the field of optical engineering and experimental solid mechanics discipline, InSAR belongs to the field of radar signal processing and mapping the Earth discipline. Although they have different in signal source, nature and physical significance, but it also need to remove high intensity speckle noise in interferogram, and extracted phase information with high precision measurement.In order to extracted more robust and accurate phase, people proposed many fringe pattern filtering methods to preserve fringe features. Such as vari-ous spin-filtering method, sparse transform method, variational regularization method and partial differential equations method. These methods consider fringe patterns’non-stationary characteristics, receive a feasible result. How-ever, it still can’t satisfied highly accuracy in actual requirements.Fringe pattern have two important characteristics:various changes in direction and frequency of different pixels. By detecting various changes and adaptive image matching, this article get robust filter results for fringe pattern.This paper conducted innovative research work in the following two aspects:1. Improved Goldstein Filtering Based on Fringe Statistics. In-SAR interferometric phase have2π hopping which leads to discontinuous, the phase-field data is easy to interference by non-stability noise, these will give difficulties in various filtering methods. First, in order to make the data con-tinuity we use cosine transform for the original image. Second, according to sparse features of phase diagram, this paper proposes an adaptive weighting coefficient instead of the original fixed values, so that the strength of the adap-tive filter could adjust according to the intensity of the noise, which get a better results than the original Goldstein filter;2. Coherence Enhancing Diffusion Based on Geometric Feature. The accurate estimate for direction of fringe pattern has a great influence for filtering method. Coherence enhancing diffusion filtering is used characteristic parameter in data tensor field directed diffusion in local areas. However, this method uses Gaussian filter calculation fringe pattern tensor field which leads to make error in fringe direction estimate. In order to obtain a accurate fringe direction and adaptive directed diffusion, we use anisotropic diffusion to make robust estimation for fringe patterns tensor field, and achieve a better filtering result finally;This paper presents two feature adaptive filterings for ESPI and InSAR patterns which better enhance image features and helpful for extracting phase field and accurate measure. These have important significance both in theory and practice.