Wavelets and curvelets transform for image denoising to damage identification of thin plate

• Published in: Results in engineering 2023-03, Vol.17, p.100837, Article 100837
• Main Authors: Yulong, Deng, Ke, Ding, Chunsheng, Ouyang, Yingshe, Luo, Yu, Tu, Jianyi, Fu, Wei, Wang, Yaguang, Du
• Format: Article
• Language: English
• Subjects: Curvelet transform; Damage identification; Image denoising; Thin plate; Wavelet transform
• ISSN: 2590-1230; 2590-1230
• DOI: 10.1016/j.rineng.2022.100837

As a common structural form, thin plates are widely used in civil engineering. Since the thin plate needs to face harsh work conditions, the damage inevitably to be accumulated, thus affecting the stability and safety of the application components. Therefore, it is of great application significance to quantify and characterize the damage of thin plates. However, the raw images produced by current inspection techniques such as Ultrasonic immersion C-scan technology, Metal sheet Lamb wave inspection technology, etc applied to thin plates usually bring various noises and imperfections during the reception, encoding, and transmission. In this paper, wavelet transform and Curvelet transform are used to denoise the detected noise image. First, we outline the numerical implementation of two newly developed multi-scale representation systems. Curvelet transform is a new multi-scale transform based on wavelet transform after 1999. The purpose of this paper is to analyze the influence of wavelet and Curvelet transform on image denoising. These methods can also be applied to the problem of image restoration from noisy images, and the effects of denoising on images are compared. The results show that the Curvelet transform can accurately identify the damage location for the thin plate damage degree, damage range, strip damage, and multiple damage conditions, and its energy focusing is better than that of the wavelet transform in each type of thin plate damage. •Poisson image denoising using multi-scale variance stable transform.•Compare the wavelet with the curve transform.•The simulation results show good noise removal performance.•Analysis and characterization of damage degree, damage distance and multiple damage conditions show that curvature transform is superior to wavelet transform.