Research On High Dynamic Range X-Ray Image Fusion

X-ray detection is one of the nondestructive determination technologies that can observe the internal structure of workpieces.It has been widely used in product quality testing of aerospace,defense industry and industrial inspection.When the object to be measured has an irregular shape and the internal structure is complex,the single X-ray image cannot contain all the structural information of the object to be measured due to the large thickness range of the object in the direction of irradiation and the limitation of the dynamic range of the imaging system.The variable energy X-ray scanning imaging technology proposed in recent years can obtain the complete structure information of the workpiece.With the help of information fusion technology,the quality detection of complex structure workpieces can be solved.Based on the technique of the variable energy X-ray scanning imaging,this thesis mainly focuses on the X-ray sequence image fusion of complex structure workpieces.Because the X-ray image has its uniqueness compared with the natural image,the grayscale chaos and the false edge problem often appear in the X-ray fused image.In this paper,based on the analysis of the characteristics of X-ray images,combined with the traditional image fusion theory and method,an improved X-ray image sequence fusion method is proposed.In order to avoid the problem of confusion of grayscale levels in the fused image,firstly,the relationship between grayscale values of X-ray images at different voltages is used to perform corresponding transformation of the image sequence before image fusion.Then the non-subsampled contourlet transform is used to decompose image in multi-scale and multi-direction.The weighted average strategy based on guided filter is used to fuse the low frequency coefficients to preserve the edge information of the image while making the fusion weight value more accurate.Pulse-coupled neural network can achieve high-frequency coefficients fusion while taking into account pixels and the overall characteristics of images.Finally,the fused image is obtained by inverse transform of NSCT.In this paper,the above fusion method is achieved through experiments,and compared with the traditional fusion methods.For the image fusion problem of X-ray sequences,the results show that the proposed two fusion algorithms have achieved good performance in both subjective visual effects and objective image quality evaluation indicators.