Benchmarking of De-noising Techniques for Streaking Artifacts in Industrial 3DXCT Scan Data

De-noising is one of the most important applications of image processing which has been applied to a wide variety of real world problems. De-noising allows for improving image quality in imaging modalities that are noise prone. A lot of research work has gone in to improving quality of 2D images using various de-noising techniques but new modalities of imaging such as industrial 3D X-ray computed tomography (3DXCT) have received little attention. Industrial 3DXCT scanning is used these days to acquire detailed images of the internal construction of industrial components and machinery so that defects within these can be identified nondestructively and non-intrusively. However, 3DXCT imaging is prone to artifacts and noise. One solution to the problem is to use increased number of projections which results in reduced noise but increased costs. A more cost effective solution is to de-noise the images. In this paper, we show how various de-noising techniques may be used to de-noise 3DXCT scan images acquired using a lower number of projections. We also benchmark these techniques using various picture quality measures. Our investigation shows that good results may be obtained using wavelet shrinkage and anisotropic diffusion.