DiffusionEdge is a diffusion probabilistic model designed for crisp edge detection. It addresses the challenge of generating accurate and sharp edge maps without post-processing. The model uses a diffusion approach in the latent space, enabling efficient training with limited resources. It incorporates an adaptive Fourier filter to adjust latent features in the frequency domain and applies uncertainty distillation to preserve pixel-level uncertainty during training. These techniques allow DiffusionEdge to produce crisp and accurate edge maps with fewer augmentation strategies. Extensive experiments on four edge detection benchmarks show that DiffusionEdge outperforms existing methods in both correctness and crispness. On the NYUDv2 dataset, it improves ODS, OIS (without post-processing), and AC by 30.2%, 28.1%, and 65.1%, respectively. The model's performance is validated through qualitative and quantitative results, demonstrating its effectiveness in generating high-quality edge maps without post-processing. DiffusionEdge is the first diffusion model for general edge detection and shows great potential for downstream tasks. However, it still faces challenges in efficiency, requiring further research to improve inference speed.DiffusionEdge is a diffusion probabilistic model designed for crisp edge detection. It addresses the challenge of generating accurate and sharp edge maps without post-processing. The model uses a diffusion approach in the latent space, enabling efficient training with limited resources. It incorporates an adaptive Fourier filter to adjust latent features in the frequency domain and applies uncertainty distillation to preserve pixel-level uncertainty during training. These techniques allow DiffusionEdge to produce crisp and accurate edge maps with fewer augmentation strategies. Extensive experiments on four edge detection benchmarks show that DiffusionEdge outperforms existing methods in both correctness and crispness. On the NYUDv2 dataset, it improves ODS, OIS (without post-processing), and AC by 30.2%, 28.1%, and 65.1%, respectively. The model's performance is validated through qualitative and quantitative results, demonstrating its effectiveness in generating high-quality edge maps without post-processing. DiffusionEdge is the first diffusion model for general edge detection and shows great potential for downstream tasks. However, it still faces challenges in efficiency, requiring further research to improve inference speed.