Research On Ray Image Enhancement And Segmentation

Due to the higher penetration ability of X-ray, X-ray has been widely used in industrial nondestructive testing. Aiming at the problems of industrial radiographic image processing and detection, some concrete researches are shown below:This thesis mainly studies the image enhancement method based on the Partial Differential Equation (PDE). At present, the existing methods still have some problems in theory and practical application, according to that, we demonstrate the improved method based on Partial Differential Equation. Stratify the image firstly and process the image in each tier by local histogram equalization and nonlinear filtering, finally, composite the image. This method overcomes the insufficiency in the classical algorithm which has not only enhanced image but also enhanced noises. The new algorithm can remove the noise effectively, and in the meanwhile, retains target details and edge well.The third chapter discuses image segmentation method based on partial differential equation, and launches research in allusion to geometric active contour model, its main content includes: (1)Geodesic active contour (GAC) model algorithm is complexity, and the large amount of calculation result in active contour evolution time long. (2)In practical applications, evolution speed in the border is usually not equal to zero, it makes the first curve which arrives in images border firstly keep moving and get into internal of the target; Or when the object image has deep sag border, the curve keep the minimum state in a local park which is different from the object's boundary. To solve these problems, this paper uses multi-scale tensor to diffuse the stop velocity field, and divides it by the GAC model. The experimental results show that this method obviously improved such flaws and the segmentation effect is significant.Based on the first two steps of image enhancement and segmentation, the defects detection method which this paper adopted is the improved mean normalization cross-correlation algorithm. According to the original algorithm with large amount of calculation and difficult for smaller defect detection problems, this paper first use scatter matrix to obtain richer information on local structure, and then extract the image edge, the edge points of the correlation coefficients is calculated, all these are done to determine that the location of the maximum correlation coefficient is the match position. This method greatly reduces computational complexity, and has high matching rate, also achieved good matching computational complexity, and has high matching rate, also achieved good matching effect.