Research On Multi-energy X-ray Image Fusion Method

Digital radiography is currently the main tool for quality inspection of complex structural devices in industry.These complex structural devices are often composed of many different materials,resulting in significant differences in equivalent thickness in the transmission direction,but the limited dynamic range of X-ray detectors and the use of single energy imaging can result in simultaneous overexposure and underexposure,making internal defects difficult to detect and thus affecting image quality and subsequent observations.For this reason,it is necessary to use multi-energy X-ray image fusion methods to solve the problem that a single energy cannot present complete information about complex structural devices.In order to achieve complete imaging of industrial complex structural devices while maintaining their original hardware conditions,this thesis employs edge-preserving filtering in multi-energy X-ray image fusion.The main research contents are as follows:1.An image fusion method based on Multiresolution Singular Value Decomposition(MSVD)and Pulse Coupled Neural Network(PCNN)is proposed.Firstly,the source image is decomposed using a gradient domain guided filter to obtain a base layer and a detail layer.Secondly,for the base layer,singular value decomposition is used to decompose it into an approximate part and three detail parts.The approximate part and the detail part are fused using a weighted average and an absolute value method,respectively.The fused two parts are reconstructed through MSVD to obtain the fused base layer image.For the detail layer,the local entropy of the source image is used as the input of the pulse coupled neural network,and the resulting output is used as the basis for detail level fusion.Finally,the fused base layer and detail layer images are reconstructed to obtain a fused image.The experimental results show that the method has a greater advantage in representing image details and can obtain fused images with clear textures.2.An image fusion algorithm based on a combination of Guided Filtering(GF),Weighted Least Squares Filtering(WLS)and Visual saliency map(VSM)is proposed.Firstly,the multi-scale image detail information is extracted by two different filtering methods,linear and non-linear,where the detail layer is the extracted detail information and the base layer is the original image.Secondly,a visual saliency map is constructed according to the principle of phase coherence and the principle of pixel contrast,from which an initial weight map is obtained and noise reduction is applied using a bootstrap filter.The weight map is then normalized to obtain the weighted mapping weights that control the fusion of the base and detail layers.Finally,the fused base and detail layers are used to reconstruct the final fusion result map.After experimental comparison,it is demonstrated that the present method can produce fused images with clear edges and good visual effects.