To address the issue of incomplete correction of coupled artifacts in cone-beam computed tomography (CBCT) images, a coupled artifact correction method for CBCT images based on a multi-scale generative adversarial network (GAN) is proposed. Firstly, a CBCT coupling artifact dataset comprising both simulated and real images was constructed based on the artifact characteristics of CT images to enhance the model′s generalization capability. Additionally, the generator structure of the network was improved by integrating the feature pyramid network (FPN) and convolutional block attention module (CBAM) to capture more comprehensive feature information. We also employed a multi-scale discriminator (MSD) alongside these components to from a generative adversarial network framework, producing clearer and more realistic artifact-free images. Experimental analysis showed that the PSNR and SSIM of the corrected images increased by this method increased by 21.595 dB and 0.541 in the simulated dataset, and by 14.072 dB and 0.274 in the real dataset. The experimental results indicate that the proposed method can effectively correct coupled artifacts.