In the Saddle: Chasing Fast and Repeatable Features

TL;DR

This paper introduces a novel saddle-like feature detector that efficiently identifies stable, repeatable points in images, outperforming similar-speed detectors in matching tasks across various challenging datasets.

Contribution

The paper presents a new similarity-covariant saddle feature detector that is fast, robust, and effective across diverse images, with a simple geometric verification method.

Findings

Saddle features are evenly distributed and dense in images.

The detector is among the fastest available.

Saddle features outperform similar-speed detectors in matching performance.

Abstract

A novel similarity-covariant feature detector that extracts points whose neighbourhoods, when treated as a 3D intensity surface, have a saddle-like intensity profile. The saddle condition is verified efficiently by intensity comparisons on two concentric rings that must have exactly two dark-to-bright and two bright-to-dark transitions satisfying certain geometric constraints. Experiments show that the Saddle features are general, evenly spread and appearing in high density in a range of images. The Saddle detector is among the fastest proposed. In comparison with detector with similar speed, the Saddle features show superior matching performance on number of challenging datasets.