Overview
GrabCut alternates between estimating foreground/background color distributions and solving a graph-cut labeling problem.
ExtendedSource checked
GrabCut
Segments foreground from background using graph cuts and color models.
Quick Facts
Compact metadata for implementation and discovery.
- Category
- Image Processing
- Difficulty
- Advanced
- Family
- Image Segmentation
- Tier
- Extended
- Status
- Verified
- Review
- Source Checked
- Sources
- 2
- Formulas
- 1
- Relationships
- 1
- Code example
- Ready
- Visualization
- Ready
Core Concepts
foreground modelbackground modelgraph cutiterative labeling
Quick facts
Complexity
Runtime and memory behavior at a glance.
Best
O(pixels)
Average
O(iterations * graph_cut)
Worst
O(iterations * graph_cut)
Space
O(pixels)
When to Use
- Use when image enhancement needs foreground model.
- Use when the input matches the assumptions behind image Segmentation.
When Not to Use
- Avoid when image noise, lighting, or scale breaks the algorithm's assumptions.
- Avoid when a simpler algorithm gives the required result with less setup.
What It Does
- Segments foreground from background using graph cuts and color models
- Uses foreground model and background model to guide the work.
- Produces processed image.
Use Cases
- Image enhancement
- Feature extraction
- Segmentation and motion analysis
Inputs
- Image or image sequence
- Kernel, threshold, or model parameters
Outputs
- Processed image
- Mask
- Feature map or motion field
Key Formulas (1)
Segmentation energy
E(labeling) = datacost + smoothnesscost
E(labeling) = data_cost + smoothness_costBalances color-model fit with boundary smoothness.
Pseudocode
- 1.Prepare the inputs and state needed for GrabCut.
- 2.Apply the foreground model step while the stopping condition has not been met.
- 3.Update intermediate state and discard invalid candidates.
- 4.Return processed image.
Canonical Example
- Input
- GrabCut input with a small representative case
- Output
- Processed image
- Explanation
- GrabCut applies foreground model to produce processed image.
Implementation Notes
- Handle borders and pixel coordinate conventions consistently.
- Keep the foreground model step separate enough to test independently.
Common Pitfalls
- Applying operations in the wrong color space can distort the result.
- Forgetting to check edge cases can make GrabCut look correct on only ideal inputs.
Interactive Mechanism Walkthrough
Spatial Primitive & Sweep-Line CanvasShow the source image beside masks, filters, or motion vectors produced by the algorithm.
Step 1 of 4: Prepare the inputs and state needed for GrabCut.
image-processing-mechanismNotes
GrabCut should be used with the assumptions and tradeoffs of image Segmentation in mind.
Sources (2)
What each reference supports for this entry.
GrabCut
PaperOriginal GrabCut reference for iterative graph-cut foreground extraction and border matting.
Open sourceComputer Vision: Algorithms and Applications
BookComputer-vision reference supporting GrabCut feature extraction, image registration, segmentation, or multiscale image-analysis workflow.
Open sourceExplore Next
Good next jumps from the relationship graph, ranked for useful discovery.
Relationships
View full relationship pageAlternative toWatershed Segmentation
GrabCut and watershed segmentation are interactive or region-based alternatives for image segmentation.
Related Algorithms
Tags
Image ProcessingSegmentationGraphOptimization