Defect Detection and Contrast Improvement of Radiographic Images using Nonlinear Diffusion Filter Method

Image prosseing methods is useful for extraction of the defects such as porosity, slag and crack in the noisy and low contrast industrial radiography images. It is noticeable that the defects appear as the gradients in pixel values, therefore, the gradient based method can help to detect the defects. In this research, different methods of gradient extraction such as Soble, Perwitt and the nonlinear diffusion filter were implemented and tried to improve the edges and the defect detection. The nonlinear diffusion filter method depends on diffusion coefficient, k, and the chosen value for k can affect the exact smoothing of the edges. Thus, in this attempt, by the use of a new approach, in comparison with the conventional edge finder methods, a nonlinear diffusion filter method is repeated and the edges are smoothed in the obtained images. Then, the defects in the gradient of images are extracted by substracting the original image from the obtained image. The mentioned methods were implemented in different radiography images. The results show that the proposed method in comparison with local gradient ones such as Sobel and Perwitt methods has superiority over different radiographic images as confirmed by the radiography film interpreters.