Santiago Betelú

3Machine learning papers tracked by Lacuna
Is this you?

Research Overview

Santiago Betelú focuses on computer vision, shape representation, and affine invariant geometry. His trajectory centers on developing robust mathematical frameworks for analyzing planar curves and three-dimensional surfaces without the need for prior denoising. The most recent work in his observed publication history addresses area-based medial axes and their applications in detecting affine skew symmetry.

Research Trajectory

Santiago Betelú’s research trajectory is characterized by a focused effort to improve the mathematical robustness of shape representation in computer vision. A central problem in early shape analysis was the extreme sensitivity of skeletal representations to noise, which typically required aggressive curve denoising prior to computation. In 2000, Betelú led work to address this limitation by introducing a novel geometric approach. In Noise-Resistant Affine Skeletons of Planar Curves, he and his collaborators proposed a new definition for affine invariant skeletons. Rather than relying on standard Euclidean or affine distances, which are prone to instability, this method defined the skeleton using the areas between a curve and its corresponding chords. This area-based distance metric proved highly robust, allowing for the direct computation of skeletons from noisy data without the need for preprocessing.

Following this foundational work on planar curves, Betelú extended these area-based geometric principles to three-dimensional surfaces. As the first author of Affine Invariant Erosion of 3D Shapes (2001), he introduced a definition of affine invariant erosion that replaced traditional Euclidean distance metrics with the volumes enclosed between a 3D surface and its chords. This construction inherently preserved affine invariance while maintaining the noise insensitivity demonstrated in his earlier 2D work. The research established key mathematical properties of this erosion operation and provided a practical method for computing the erosion of implicit surfaces, laying the groundwork for the extraction of 3D affine invariant robust skeletons.

Betelú continued to refine these area-based representations in subsequent collaborative research. In Area-Based Medial Axis of Planar Curves. (2004), where he served as second author, the research formalized the affine invariant medial axis for closed planar curves. This work provided a dynamical interpretation of the medial axis, demonstrating that medial axis points correspond exactly to the affine shock positions of the curve's affine erosion. The methodology was successfully applied to real-world image analysis, specifically demonstrating its utility in the detection of affine skew symmetry.

Current Directions

The most recent paper in the provided evidence is from 2004, and it points toward a broader interest in the dynamical interpretation of shape evolution and symmetry detection. This direction focuses on linking static geometric representations, such as the medial axis, to dynamic mathematical processes like affine erosion and shock formation. By formalizing the relationship between area-based distance metrics and affine shock positions, this research addresses the challenge of extracting stable, invariant features from real images. The primary application of these methods lies in detecting affine skew symmetry under noisy conditions, providing a purely geometric alternative to traditional, noise-sensitive distance transforms in computer vision.

Profile updated

See something wrong? Report a mistake