Research On X-ray Multi-spectrum Blind Separation Algorithm Based On Statistical Prior

Compared with the traditional single energy CT image,the multispectral X-ray CT image has more abundant spectral information,and can effectively distinguish the composition and attenuation coefficient of the detected object.In this paper,the energy spectrum separation method of variable energy X-ray projection sequence is studied without changing the physical composition of the traditional CT imaging system,so as to realize the narrow spectrum multi-energy CT image imaging,in order to effectively distinguish the material.And the quantitative characterization of microstructure provides support.Firstly,the theoretical basis of X-ray imaging is analyzed,and combined with the attenuation mechanism of X-ray,it is analyzed that multispectral X-ray with variable voltage can be decomposed by non-negative matrix.By using the statistical model of projection data and the non-negative matrix decomposition method,the optimization model is established,combined with KKT optimization conditions,the model is optimized and improved,and the established model is deduced and solved,and the separation of energy spectrum sequences is realized.The corresponding narrow energy spectrum information is obtained.In order to verify the effectiveness of the algorithm,this paper simulates the modules containing two kinds of silicon and aluminum with similar attenuation coefficients,and obtains their projection sequences at three voltages by using the calculation proposed in this paper.Eighteen narrow-spectrum projections were obtained by decomposition,and silicon and aluminum were successfully zoned.