Research On Point Spread Degradation Modeling And Image Restoration For Digital Radiography

Point spread function(PSF) is an important indicator to characterize response property of the X-ray digital radiography system, its diffusion form directly determines the degree of degradation of DR images. In order to restore and improve the quality of DR degraded images, following researches were launched for digital radiography system based on GE TITAN E320 X-ray generator and PerkinElmer XRD0822 flat panel detector.Firstly, edge spread functions(ESF) and line spread functions(LSF) under different angles were measured by using rotatable edge measurement method, then the reconstruction data of point spread distribution was obtained from LSF data with filtered-back projection algorithm. Its accuracy indirectly verified by MTF measured with bar-pattern test object. On these bases, a model for point spread distribution data was proposed, the model fitting test indicates that adjusted R-square value of the proposed model for reconstructed point spread distribution data reaches more than0.999 and the fitting residuals of points were only between-0.015 and 0.01, which has a greatly improved degree of compliance compared with traditional PSF model of the Gaussian function. Simultaneously, the change of diffusion shape of LSF and PSF affected with tube voltage, tube current and radiographic magnification was studied and several conclusions can be drawn from this research: at the range from 120 kV to200kV, the diffused extent of PSF and LSF at each angle slight increase with tube voltage; in the range of 1mA to 4mA, the diffused extent of PSF and LSF at each angle are irrelevant with tube current; within 1.5 to 3.5 times magnification,the diffused extent of PSF and LSF at each angle increase with radiographic magnification significantly, which indicates the diffusion form of PSF is greatly affected by the geometric unsharpness in X-ray digital radiography system.Secondly, the sparsity of noise component and texture component of DR image in the fourier domain and wavelet domain were studied respectively by the image sparse representation test. Test results show that fourier domain has a large advantage over wavelet domain in sparse representation for DR noise signal but less sparse than wavelet domain in texture representation. Either of them can not give sparse representations for both noise and texture in DR image.Based on researches and conclusions above, a restoration algorithm for DRimage combining fourier domain and wavelet domain was proposed and its estimation error in restoration was reduced by balanced parameter optimization. Further,restoration tests were conducted by using the method of measuring point spread distribution in this paper and the proposed restoration algorithm. Tests and quality evaluation for results show that image restoration algorithm proposed in this paper can effectively deblur degraded DR images by preserving DR image texture to the greatest extent while suppressing noise, which has a outstanding performance on restoration of imaging edge structure, texture details, defects and distinct advantage in the subjective and objective evaluation compared with Wiener filter and WWF algorithm, which is based on a single transform domain. Hence, it can be well applied to X-ray digital radiographic technology.