Processing Method Of Metal Fracture Images Based On Grouplet Transform

This thesis is supported by the national natural science foundation of China (No.51261024) and Ministry of Education Research (No.ZD200829003). Grouplet transformis applied to image preprocessing of the metal fracture. Considering the problems of thedecline of the metal fracture image quality, using Grouplet transform for edge detection,image denoising and image enhancement processing, made some innovativeachievements. The main content of this paper includes the following several aspects:The first chapter discusses the issues background and research significance,reviews the research status at home and abroad of the fracture image pretreatment andthe Grouplet transform, and puts forward the main content and innovation of this paper.The second chapter discusses some basic concepts and basic theory about Grouplettransform, expounds its advanced than ordinary wavelet transform. Then combiningwith the defect of Grouplet transform and the advantages of Bandelet transform, comesup with Grouplet-Bandelet transform theory.The third chapter discusses the traditional edge of the first order differentialoperator and the second order differential operator, and the wavelet edge detectionalgorithm. And aiming at the shortcomings of the commonly used edge detectionalgorithm, using the Grouplet transform coefficient can represent any natural imagetexture direction of geometric characteristics, proposes new edge detection methodbased on Grouplet transform modulus maxima. Through the experimental study showsthat the new method can accurately extract edge banding of metal fatigue fracture image,and this is very helpful for the precise calculation of fatigue banding cycle. However,LOG operator, Canny operator and edge detection operator based on wavelet transformto detect fatigue strip edge position is not accurate, obviously appear more false edge, isnot conducive to precise calculation of the fatigue banding cycle.The fourth chapter introduces the characteristics of image denoising based onwavelet transform and its defects in the image representation. And aiming at theshortcomings of the wavelet transform denoising, puts forward the new denoisingmethods based on the Grouplet transform and Grouplet-Bandelet transform. Throughsimulation and experiment research,it proves that the denoising method based onGrouplet transform performed better than the wavelet transform denoising algorithm.The advantages mainly displays in three aspects: image clarity, increasing of the peaksignal to noise ratio and protecting image detail texture, however, these advantages arenot obvious. The denoising method based on Grouplet-Bandelet transform is better thanthe denoising methods based on single Grouplet transform and Bandelet transform ondenoising effect.The fifth chapter illustrates the advantages and disadvantages between thetraditional image enhancement algorithms and image enhancement algorithm based on wavelet transform, puts forward an image enhancement algorithms based on Grouplettransform and Grouplet-Bandelet transform.The simulation and experimental studiesshow that the image enhancement algorithms based on Grouplet transform is muchbetter than the wavelet transform in enhancing fatigue stripe edge of fatigue fractureimage. The image enhancement algorithms based on Grouplet-Bandelet transform is notas good as Grouplet transform in enhancing the image texture edge, but it can eliminateimage noise to a certain extent, so that the enhanced image has higher contrast and hasno distortion.The sixth chapter has made a detailed summary of this paper and puts forwardsome questions which is worth to study in the future.