A Hybrid Image Restoration Method based on Partition for Industrial X-ray images

Abstract

The balance between image denoising and detail-maintaining is a significant index of image restoration. X-ray images of integrated circuits (IC) are characterized by high level of noise and low contrast, so that tiny details preserving is one of the key points to the internal defect detection of IC. Therefore, a hybrid image restoration method is proposed. First of all, by using Fourier transform-based image partition method, restored results of iterative denoising and backward projections (IDBP) and adaptive total generalized variation method (ATGV) are divided into detail areas and smooth areas. Then, smooth area of IDBP and detail area of ATGV are combined to obtain a new restored X-ray image, which can make full use of the excellent denoising effect of IDBP and the detail-preserving performance of ATGV. Experimental results of X-ray images show that the proposed method has the ability to attain the balance between denoising and detail-maintaining.