Anisotropic Diffusion
Reduce image noise with Perona-Malik anisotropic diffusion while preserving edges and important structural boundaries.
Perona-Malik anisotropic diffusion smooths flat regions while preserving sharp edges. Ideal for denoising while keeping important structure.
How to Use Anisotropic Diffusion
- Upload an image: Upload one of the formats listed by the tool (JPG, PNG, WebP) and wait for the local preview.
- Adjust settings and process: Set the number of iterations, diffusion constant (kappa), and time step (delta) to control the smoothing behavior.
- Review and download: Review the result, then download it in one of the formats offered by this tool.
Frequently asked questions
What is anisotropic diffusion?
Anisotropic diffusion is an edge-preserving smoothing technique based on the heat equation. Unlike Gaussian blur which smooths everything uniformly, it adapts the diffusion rate based on local gradients. Smoothing flat areas while leaving edges sharp.
How is it different from bilateral filtering?
Both preserve edges, but anisotropic diffusion is an iterative PDE-based method that naturally evolves over time, while bilateral filtering is a single-pass weighted average. Anisotropic diffusion often produces more natural-looking results for denoising.
How is anisotropic diffusion different from bilateral filtering?
Both preserve edges while smoothing. Bilateral filtering uses a single-pass weighted average based on spatial and color distance. Anisotropic diffusion is an iterative PDE-based process that progressively smooths over multiple steps. Bilateral is faster; anisotropic diffusion gives finer control over smoothing strength.
Is the Anisotropic Diffusion tool free?
Yes. All tools on Pixlane are free with no signup required.
Does it reuse cached assets on repeat visits?
Browser caching can speed repeat visits; uncached site assets can still require an internet connection.
What image formats are supported?
JPG, PNG, and WebP are supported.
What do the alpha and K parameters control?
Alpha controls the diffusion rate per iteration. Higher values smooth more aggressively (must stay below 0.25 for numerical stability). K is the gradient threshold: edges with gradients above K are preserved. Lower K preserves more edges; higher K smooths more aggressively.
How many iterations should I use?
Start with 5-10 iterations for light denoising. Use 20-30 for stronger smoothing. More iterations progressively smooth flat regions while maintaining edges, but beyond 40-50 the effect plateaus and processing slows.