Optimized CT Image Denoising Method Combining Wavelet Transform and U-Net Model

Low-dose computed tomography (CT) is essential for minimizing patient radiation exposure; however, increased noise often leads to image degradation and may adversely affect diagnostic accuracy. In this study, we propose an optimized CT image restoration method that integrates wavelet transform with the U-Net architecture. The proposed approach decomposes the input image into low- and high-frequency components, selectively removes noise from the high-frequency bands, and reconstructs the image while preserving structural information in the low-frequency bands. The reconstructed components are subsequently refined through the U-Net for final denoising. Performance was quantitatively evaluated using PSNR and SSIM, showing improvements in the average PSNR by 10.3% and average SSIM by 14.7%, respectively, compared to the conventional U-Net. These results demonstrate that the wavelet-based U-Net model offers superior denoising performance while maintaining image resolution and anatomical structure, suggesting it as an effective approach for improving the quality of low-dose CT images.