The Research Of De-noising Algorithm In Active Thermal Imaging Based On Scale Transform

Affected by environment,infrared thermal imager and surface characteristics of samples,there are a large of noise in active thermography images.It is not conducive to the detection of defects,so we need to filter the noise in the active thermal images by de-noising algorithm.In view of the fact that there are few studies on the active thermal imaging de-noising algorithm,shearlet transform is chosen as scale transform method which is used in de-noising algorithm,based on the studies of passive thermal imaging de-noising methods.An image de-noising algorithm based on adaptive threshold and inter-scale coefficient variation is proposed by analyzing the characteristic of the non-subsampled shearlet transform of the pulsed eddy current thermography image.The effect of de-noising algorithm is test by using different pulsed eddy current thermography images,and the visual effect,the noise standard and the signal-noise ratio are used as evaluation indexes.The experimental results show that compared with WBC adaptive threshold de-noising algorithm,Contourlet adaptive threshold de-noising algorithm and other traditional de-noising algorithm,the proposed algorithm has better de-noising effect.The specific research contents of this paper are as follows:(1)The noise characteristics of infrared thermal imager were studied by frame difference method,and it was found that noise with different frame interval can be described by generalized Gaussian distribution model.The noise characteristics of pulsed eddy current thermography images were studied by multi-frame images superimposed on average method,and it was found that the noise type of image is Gaussian noise.(2)The noise sources of active thermal imaging detection were introduced.In order to enhance the signal and prevent signal data from being damaged by de-noising algorithm,the method of image feature extraction step was done before de-noising step was put forward,based on the particularity of active thermal image data;The advantages and disadvantages of spatial domain de-noising method and transform domain de-noising method were compared,to verify the correctness of using transform domain de-noising method in this paper.The characteristic of each scale transform method was studied,and non-subsampled shearlet transform was chosen to use in the proposed de-noising algorithm in this paper.(3)The commonly used threshold function and threshold expressions were described,and an adaptive threshold expression related to scale was proposed based on the universal threshold expression.For pulsed eddy current thermography images,it was found that the noise coefficients and signal coefficients can be distinguished well by root mean square difference(RMSD)by analyzing the variation characteristics of non-subsampled shearlet transform between scales.Therefore,a pulsed eddy current thermography image de-noising algorithm based on adaptive threshold and inter-scale coefficient characteristic was proposed.(4)Pulsed eddy current thermography images with non-crack,0.5mm crack and 2.0mm crack were used to test the effect of de-noising algorithm in this paper.After the de-noising results of the proposed algorithm were compared with other algorithm's results,the de-noising effectiveness for pulsed eddy current thermography images,the advantages of non-subsampled shearlet transform against other scale transform method,the advantages of distinguishing noise and signal by adaptive threshold and inter-scale coefficient characteristic were verified.