PCB Defect Images Super-Resolution Reconstruction Based on Improved SRGAN

Image super-resolution reconstruction technology can boost image resolution and aid in the discovery of PCB flaws. The traditional SRGAN algorithm produces reconstructed images with great realism, but it also has disadvantages insufficient feature information extraction ability, a large number model parameters, as well lack fine-grained impact. To that end, this paper proposes an SRGAN-based for defect is first to add VIT network generation extend perceptual field improve model’s ability extract high-frequency information. module then used enhance generator’s from map while reducing complexity network. Finally, inverted residual are combined form discriminator’s backbone network, which extracts summarizes shallow features synthesizing global higher-level features, allowing discriminator effect be achieved less spatial complexity. On test set, improved increases PSNR by 0.82 SSIM 0.03, SRVIT algorithm’s parameters space size decreased 2.01 M 7.5 MB, respectively, when compared algorithm. Moreover, not only enhances lowers