Wavelet thresholding Genetic algorithm approach for Noise Extraction in High-Resolution Industrial Tomography

Abstract

Since the development of Computed Tomography (CT) in medicine, many applications have been emerging among which the use in Non-Destructive Evaluation (NDE) approach has been consolidating in recent years for analysis of inner features in a broad range of industrial components. More recently, this method has also been applied fordimensional measurements in the metrology field. During acquisition stage many artifacts may cause distortions that interfere with the sample edge evaluation, thereby generating errors on the surface determination. In such development, high accuracy is required for its use in metrology and overall volumetric reconstruction. Scatter radiation is a major concern in the image acquisition process, being strongly dependent on the object densities and geometry. A combined approach involving genetic algorithm and wavelet shrinkage is proposed for denoising application, where 2D radiographic projections are filtered prior to the volumetric reconstruction process. The developed algorithm is applied to sample images resulting from tomography procedures that usually produce severe artifacts and is evaluated in terms of Peak Signal-to-Noise Ratio (PSNR). The filtering technique advanced in this paper generates reconstructed volumes with less noise, accurate edges and improved visual perception.