Graph-Cut RANSAC: Local Optimization on Spatially Coherent Structures.

Area of Science:

• Computer Vision
• Geometric Modeling
• Optimization Algorithms

Background:

• Robust geometric model estimation is crucial for many computer vision applications.
• Traditional methods like RANSAC can struggle with high levels of noise and outliers.
• Existing advanced methods may lack efficiency or optimal inlier selection.

Purpose of the Study:

• To introduce Graph-Cut RANSAC (GC-RANSAC), a novel robust method for geometric model estimation.
• To leverage graph-cut for energy minimization and optimal inlier selection.
• To improve accuracy, reduce failure rates, and maintain efficiency in geometric estimation.

Main Methods:

• Formulating the local optimization step as energy minimization with binary labeling.
• Applying the graph-cut algorithm for inlier selection based on spatial coherence.
• Incorporating unary terms for point-to-model residuals and binary terms for neighbor labeling.

Main Results:

• GC-RANSAC demonstrates superior geometric accuracy compared to state-of-the-art methods.
• The method exhibits a lower failure rate across various estimation tasks.
• Achieves competitive or faster execution speeds than less accurate alternatives.

Conclusions:

• GC-RANSAC provides a conceptually simple, efficient, and globally optimal inlier selection.
• The method is effective for diverse geometric estimation problems like homography and 6D pose.
• The open-source availability facilitates further research and application.