CCD-DR Image Calibration

Digital x-ray image technology is one of the most active research field of medical compared with the traditional x-ray imageing technology. Characterized by faster image acquisition, higher image resolution and higher radiation efficiency, it is important for medical image analysis and diagnosis. In recent years, superior performance of DR is taken seriously by local enterprises with the support of the department.Due to the defect of present mechanical manufacturing process and the influence of the external environment, grayscale distortion and geometry distortion is inevitable in CCD-DR system. Image details disappear and visual perception is not real which go against analysis and diagnosis of doctor. Considered the actual demand of the enterprise and the development costs, effective algorithm is put forward based on the analysis of factors which lead to the decline of image quality.To solve the problem of gray correction, dark current noise and the non-uniformity of the x-ray image are focused. After comparing several common correction methods with experiments, the piecewise linear fitting algorithm with the highest correction accuracy and least space cost is used to deal with the gray correction. When the x-ray is excessively used, the assumed conditions of piecewise linear fitting algorithm cannot be met, based on which the author proposes regional compensation algorithm. The contrast of images before and after correction verifies the feasibility and effectiveness of the proposed algorithm.To solve the problem of geometrical correction. A phantom with suitable geometrical properties, which consists of organic glass and cladding steel sheet, is designed to calibrate the geometrical parameters of the correction model. An algorithm based on DRLSE algorithm is put forward to extract the identification points in the phantom. Polynomial transform method, orthogonal polynomial transformation method, moving least square method are used for geometrical correction.Based on the accuracy in calibration, moving least square method is proved to be more accurate. This algorithm has been proved to be valid and effiective.Acceleration algorithm need to be designed to meet real-time needs of radiography. Thus, the study finally applies the Nvidia’s CUDA technique to the piecewise linear fitting algorithm and geometrical correction method. Design the parallel computing architecture and finally the execution speed of the proposed algorithms is improved.