# Image Denoising

γ-photon flow-field images reconstructed from limited coincidence events have blurred boundaries and local structures that noise can swamp; a conventional conditional diffusion model injects the degraded image only by concatenation, so the structural prior is diluted through repeated downsampling, and its frequency processing does not change with the diffusion step. The Structure-Frequency Guided Diffusion (SFGD) model feeds edge and morphology priors into the denoiser throughout via a multi-scale structure stream and windowed cross-attention, and tunes the range and gain of frequency enhancement by diffusion timestep. On a phantom-level independent test set, the full model reaches 28.325 dB PSNR and 0.926 SSIM, 5.499 dB and 0.153 above the conditional diffusion baseline.

γ-photon flow-field images reconstructed from limited coincidence events have blurred boundaries and local structures that noise can swamp; a conventional conditional diffusion model injects the degraded image only by concatenation, so the structural prior is diluted through repeated downsampling, and its frequency processing does not change with the diffusion step. The Structure-Frequency Guided Diffusion (SFGD) model feeds edge and morphology priors into the denoiser throughout via a multi-scale structure stream and windowed cross-attention, and tunes the range and gain of frequency enhancement by diffusion timestep. On a phantom-level independent test set, the full model reaches 28.325 dB PSNR and 0.926 SSIM, 5.499 dB and 0.153 above the conditional diffusion baseline.

This study proposes HARU-Net, a hybrid attention residual U-Net trained on a cadaver dataset of human hemimandibles acquired with a high-resolution cone-beam computed tomography (CBCT) protocol for low-dose CBCT denoising; the architecture embeds a hybrid attention transformer block within each skip connection, a residual hybrid attention transformer group at the bottleneck, and residual learning convolutional blocks, and reports the highest peak signal-to-noise ratio of 37.52 dB, the second-highest SSIM of 0.9557, and the lowest GMSD of 0.1084, while maintaining substantially lower computational complexity than transformer-based methods.

This study proposes HARU-Net, a hybrid attention residual U-Net trained on a cadaver dataset of human hemimandibles acquired with a high-resolution cone-beam computed tomography (CBCT) protocol for low-dose CBCT denoising; the architecture embeds a hybrid attention transformer block within each skip connection, a residual hybrid attention transformer group at the bottleneck, and residual learning convolutional blocks, and reports the highest peak signal-to-noise ratio of 37.52 dB, the second-highest SSIM of 0.9557, and the lowest GMSD of 0.1084, while maintaining substantially lower computational complexity than transformer-based methods.