Abstract:In this paper, we point out that suboptimal noise-data mapping leads to slow training of diffusion models. During diffusion training, current methods diffuse each image across the entire noise space, resulting in a mixture of all images at every point in the noise layer. We emphasize that this random mixture of noise-data mapping complicates the optimization of the denoising function in diffusion models. Drawing inspiration from the immiscibility phenomenon in physics, we propose Immiscible Diffusion, a simple and effective method to improve the random mixture of noise-data mapping. In physics, miscibility can vary according to various intermolecular forces. Thus, immiscibility means that the mixing of molecular sources is distinguishable. Inspired by this concept, we propose an assignment-then-diffusion training strategy to achieve Immiscible Diffusion. As one example, prior to diffusing the image data into noise, we assign diffusion target noise for the image data by minimizing the total image-noise pair distance in a mini-batch. The assignment functions analogously to external forces to expel the diffuse-able areas of images, thus mitigating the inherent difficulties in diffusion training. Our approach is remarkably simple, requiring only one line of code to restrict the diffuse-able area for each image while preserving the Gaussian distribution of noise. In this way, each image is preferably projected to nearby noise. Experiments demonstrate that our method can achieve up to 3x faster training for unconditional Consistency Models on the CIFAR dataset, as well as for DDIM and Stable Diffusion on CelebA and ImageNet dataset, and in class-conditional training and fine-tuning. In addition, we conducted a thorough analysis that sheds light on how it improves diffusion training speed while improving fidelity. The code is available at <a class="link-external link-https" href="https://yhli123.github.io/immiscible-diffusion" rel="external noopener nofollow">this https URL</a>

Anisotropic Diffusion Based on Directional Consistency

Analysis and Design of Anisotropic Diffusion for Image Processing

Behavioral Analysis of Anisotropic Diffusion in Image Processing .

The Directional Diffusion Based on Discrete Wavelet Algorithm to Achieve Side-Scan Sonar Image Denoising

Speckle suppression using anisotropic diffusion based on information-theoretic heterogeneity measurement

Low-contrast Edge Enhancing Anisotropic Diffusion for Speckle Reduction

Image denoising by anisotropic diffusion in wavelet domain

Image Processing for Low-Dose CT via Novel Anisotropic Fourth-Order Diffusion Model

Anisotropic Diffusion Probabilistic Model for Imbalanced Image Classification

A Directional Diffusion Algorithm for Inpainting

AN ANISOTROPIC DIFFUSION EQUATION THAT CAN REMOVE NOISE AND KEEP NATURALNESS

Immiscible Diffusion: Accelerating Diffusion Training with Noise Assignment

NoiseDiffusion: Correcting Noise for Image Interpolation with Diffusion Models beyond Spherical Linear Interpolation

AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model

A fast computational technique based on a novel tangent sigmoid anisotropic diffusion function for image-denoising

The Technology Research on Image Smooth and Segmentation Based on Anisotropic Diffusion

An Adaptive Anisotropic Diffusion Filter for Noise Reduction in MR Images

Edge-preserving noise for diffusion models

3D-Consistent Image Inpainting with Diffusion Models

Speckle-Constrained Nonlinear Anisotropic Diffusion

An Anisotropic Diffusion PDE for Noise Reduction and Thin Edge Preservation