PCB image restoration method based on the Bayesian optimization TWIST algorithm
Jun Wang, Youbin Gong, Yi Wu · Journal of Physics Conference Series · 2025
Abstract Aiming at the noise in the process of image acquisition, this paper proposes a PCB image restoration model based on the Bayesian optimization TWIST algorithm. By fusing non-local mean filtering, the observation matrix and regularization parameters of the TWIST algorithm are optimized by Bayesian optimization. In this paper, 32 PCB images are selected as the experimental objects of restoration effect and defect detection, and 8 of them are randomly selected for comparison. In the experimental results of the restoration effect, compared with the original algorithm, MSE is reduced by 70.07 % of the original algorithm on average, PSNR is increased by 12.12 % on average, and SSIM is increased by 1.857 % on average. In the experimental results of defect detection, the confidence of the defect detection of the restoration model in this paper reaches 80.2 %, which is 1.4 % higher than the 78.8 % of the traditional TWIST algorithm. Experiments verify the excellent image restoration ability and practicability of the image restoration model in this paper.