Field of Junctions: Extracting Boundary Structure at Low SNR

TL;DR

This paper presents a unified bottom-up model that detects boundaries, corners, and junctions in images, especially effective at high noise levels, by analyzing local boundary structures with a novel regularizer.

Contribution

It introduces a generalized M-junction model and a non-convex optimization approach for boundary analysis, enabling simultaneous detection of contours, corners, and junctions in noisy images.

Findings

Effective boundary detection at high noise levels

Simultaneous detection of contours, corners, and junctions

Improved segmentation robustness

Abstract

We introduce a bottom-up model for simultaneously finding many boundary elements in an image, including contours, corners and junctions. The model explains boundary shape in each small patch using a 'generalized M-junction' comprising M angles and a freely-moving vertex. Images are analyzed using non-convex optimization to cooperatively find M+2 junction values at every location, with spatial consistency being enforced by a novel regularizer that reduces curvature while preserving corners and junctions. The resulting 'field of junctions' is simultaneously a contour detector, corner/junction detector, and boundary-aware smoothing of regional appearance. Notably, its unified analysis of contours, corners, junctions and uniform regions allows it to succeed at high noise levels, where other methods for segmentation and boundary detection fail.