Eddy Current Scanning Image Denoising Method Based on Principal Component Analysis and Manifold Learning

Abstract

Due to the complicated industrial environment and the poor surface conditions of detected materials, scanning images inevitably contain various noise in actual eddy current imaging detection, which seriously affects the detection result. Aiming at the above problem, we propose an eddy current scanning image denoising method based on principal component analysis (PCA) and locally linear embedding (LLE) in this paper. First, the method uses PCA to preliminarily remove noise from the scanning image. Then, the method uses the reconstruction algorithm of LLE to reconstruct the PCA-processed image by its neighborhoods, which further denoise the eddy current scanning image and optimize their details and edges while retaining their local geometric constructions. The experimental results have shown that, compared with other methods, the proposed method not only removes noise more effectively but also retains the details of the scanning image.