Defect Detection Improvement of Digitised Radiographs by Principal Component Analysis with Local Pixel Grouping

Abstract

Radiographic inspection is one of the most important techniques among non-destructive testing methods. Radiographic images are often very noisy and the image quality and the interpreter’s experience can affect the inspection of radiographs and their evaluation. In this research, principal component analysis (PCA) with local pixel grouping (LPG) algorithms was used for image enhancement for radiograph image interpretation. In this method, a pixel and its neighbors are considered as a vector variable for preservation of radiography image local structure. This method is a statistical method that uses an orthogonal property to transform and convert a set of observations of possibly correlated variables into a set of values of uncorrelated variables. Here, the PCA-LPG denoising algorithm has been applied to radiographic images with different defects to obtain denoised images. The results show that the contrast of denoised radiography images is better than the original image and the defects are much clearer. Also, the evaluation of the image quality enhancement show the contrast to noise level increases almost two times by the proposed PCA-LPG method.