Smooth

The Smooth tool applies surface smoothing algorithms to reduce noise, remove small irregularities, and improve the overall appearance of mesh surfaces. The tool offers two distinct smoothing approaches—each with different characteristics suited to particular applications.

Overview

Surface smoothing is a fundamental mesh processing operation that adjusts vertex positions to reduce local curvature variations and surface irregularities. While conceptually simple, effective smoothing requires balancing noise reduction against feature preservation and volume shrinkage.

Common applications include:

• Noise reduction: Removing scanning artifacts and segmentation irregularities
• Visual improvement: Creating smoother, more aesthetically pleasing surfaces
• Pre-processing: Preparing surfaces for downstream operations that benefit from cleaner input
• Post-processing: Refining surfaces after Boolean operations or remeshing

Accessing the Tool

Navigate to the Surface ribbon tab and locate Smooth in the Edit section. Select one or more surface objects before activating the tool.

Smoothing Methods

Two fundamentally different smoothing algorithms are available:

Basic (Laplacian Smoothing)

Laplacian smoothing is an iterative algorithm that moves each vertex toward the average position of its neighbors. This classic approach:

• Mechanism: Each vertex moves toward the centroid of its neighboring vertices
• Characteristics: Simple, predictable behavior; progressive smoothing with iterations
• Limitation: Causes volume shrinkage over many iterations

Laplacian smoothing is intuitive and controllable. Low iteration counts provide subtle smoothing, while higher counts produce increasingly smooth surfaces. However, the shrinkage effect becomes pronounced with many iterations.

Best suited for:

• Quick noise reduction with few iterations
• Situations where some shrinkage is acceptable
• Simple, predictable smoothing requirements

Smart (Windowed Sinc Smoothing)

Windowed Sinc smoothing is a frequency-based filtering method that provides superior smoothing characteristics by treating the mesh as a signal and filtering out high-frequency components (noise) while preserving low-frequency components (overall shape).

• Mechanism: Spectral filtering using a windowed sinc function
• Characteristics: Minimal volume shrinkage, better feature preservation
• Advantage: Maintains overall shape while removing noise

This method is generally preferred for quality-critical applications as it produces superior results without the shrinkage problems of Laplacian smoothing.

Best suited for:

• High-quality smoothing with shape preservation
• Anatomical models where accuracy matters
• Multiple smoothing passes without cumulative shrinkage

Parameters for Basic (Laplacian) Method

Number of Iterations

Controls how many smoothing passes are applied. Each iteration moves vertices toward their neighbors' average position:

Iterations	Effect	Typical Use
1-5	Subtle smoothing	Minor noise reduction
10-20	Moderate smoothing	General purpose
50-100	Strong smoothing	Significant irregularity removal
100+	Aggressive smoothing	Major surface simplification

Higher iteration counts produce smoother results but increase both processing time and volume shrinkage.

Relax Factor

The relax factor (0.0 to 1.0) controls how far vertices move toward their neighbors' average in each iteration:

• Lower values (0.1 - 0.3): Conservative smoothing; each step makes small adjustments
• Medium values (0.4 - 0.6): Balanced smoothing behavior
• Higher values (0.7 - 1.0): Aggressive per-step smoothing

A higher relax factor with fewer iterations can achieve similar results to a lower factor with more iterations, but the behavior at feature edges may differ.

Parameters for Smart (Windowed Sinc) Method

Number of Iterations

Similar to Laplacian smoothing, this controls the number of filtering passes. However, because Windowed Sinc preserves volume better, more iterations can be used without the same shrinkage concerns:

Iterations	Effect
5-10	Light filtering
15-25	Moderate smoothing
30-50	Strong smoothing

Pass Band Value

The pass band value controls the cutoff frequency of the filter, determining which surface features are preserved versus smoothed:

• Lower values (0.01 - 0.05): Aggressive filtering; only the largest features preserved
• Medium values (0.1 - 0.2): Balanced filtering for general use
• Higher values (0.3 - 0.5): Conservative filtering; more detail preserved

Think of the pass band as a "detail preservation" control—lower values remove more detail, higher values preserve more.

Feature Edge Smoothing

Enable Feature Edge Smoothing

When enabled, the algorithm specifically handles vertices along sharp edges differently from interior vertices. This provides control over whether sharp edges should be smoothed or preserved.

Feature Angle

The feature angle threshold (in degrees) defines what constitutes a sharp edge:

• Edges where adjacent triangles meet at angles greater than the feature angle are considered sharp edges
• Default value: 45°
• Range: 0° to 180°

Lower feature angles (15-30°): Only the sharpest edges are protected; more aggressive smoothing Higher feature angles (60-90°): More edges are considered "features" and protected from smoothing

When feature edge smoothing is enabled, vertices along detected feature edges are either:

• Constrained to move only along the edge direction, or
• Prevented from moving entirely

This preserves the character of designed sharp edges while smoothing the rest of the surface.

Practical Guidance

Choosing Between Methods

Scenario	Recommended Method
Quick preview smoothing	Basic (Laplacian)
Quality-critical output	Smart (Windowed Sinc)
Anatomical models	Smart (Windowed Sinc)
Mechanical parts with sharp edges	Either, with feature edge smoothing
Iterative refinement	Smart (Windowed Sinc)

Preserving Sharp Edges

For models with intentional sharp edges (mechanical parts, architectural models):

1. Enable Feature edge smoothing
2. Set the Feature angle to match your edge sharpness (typically 30-60°)
3. Use moderate smoothing parameters

Addressing Volume Shrinkage

If Laplacian smoothing causes unacceptable shrinkage:

1. Switch to Windowed Sinc method
2. Use fewer iterations with lower relax factor
3. Consider using the Remesh tool as an alternative

Smoothing Noisy Segmentation Results

For surfaces generated from image segmentation:

1. Start with Smart (Windowed Sinc) method
2. Use 15-25 iterations
3. Set pass band to 0.1-0.15
4. Disable feature edge smoothing (segmentation artifacts aren't true features)

Technical Considerations

Vertex Connectivity

Smoothing operates on vertex positions based on mesh connectivity. Isolated vertices or poorly connected regions may smooth differently than well-connected areas.

Boundary Handling

Open surfaces with boundary edges have special handling—boundary vertices are typically constrained to smooth only along the boundary, preserving the open edge.

Computational Cost

• Laplacian smoothing: Linear in vertex count × iterations; very fast
• Windowed Sinc: More complex per-iteration computation; moderate speed

Both methods scale well to large meshes, with Windowed Sinc requiring more computation per iteration but often fewer iterations for equivalent results.

Reversibility

Smoothing modifies vertex positions irreversibly—lost detail cannot be recovered. Always maintain a backup of the original surface when experimenting with smoothing parameters, or use the Create new surface option to preserve the original.

Common Issues and Solutions

Issue	Likely Cause	Solution
Excessive shrinkage	Too many Laplacian iterations	Switch to Windowed Sinc
Sharp edges rounded	Feature smoothing disabled	Enable feature edge smoothing
Noise not removed	Insufficient iterations/passes	Increase iterations or lower pass band
Processing too slow	Very high iteration count	Reduce iterations; use Windowed Sinc
Uneven smoothing	Non-uniform mesh density	Remesh first for uniform triangles