AI Paper Atlas

Research map / 3D vision / Depth and stereo

A Polynomial Formula for the Perspective Four Points Problem

David Lehavi, Brian Osserman

IJCV 2026 · 3D vision › Depth and stereo · approach: Camera model and calibration for SfM · 0 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use the explicit four-point formula instead of EPnPP4P solver · Runtime per configurationtime per configuration (us): 25.771 -> 0.477Improved
use the explicit four-point formula instead of SQPnPP4P solver · Runtime per configurationtime per configuration (us): 36.312 -> 0.477Improved
compile with AVX2 (ablation)P4P solver · Runtime per configurationtime per configuration (us): 0.477 -> 0.258Improved
use strict error threshold 0.05 instead of threshold 1 (ablation)Ours · General configuration, noise 0rotation error (degrees): 6.7 -> 0.5Improved
use strict error threshold 0.05 instead of threshold 1 (ablation)Ours · General configuration, noise 0success count: 9387 -> 7884Hurt
use ours (threshold 0.05) instead of EPnPOurs · General configuration, noise 0rotation error (degrees): 11.4 -> 0.5Improved
use ours (threshold 0.05) instead of SQPnP at high noiseOurs · General configuration, noise 30rotation error (degrees): 5.0 -> 6.0Hurt
use ours (threshold 0.05) instead of SQPnP on collinear triplesOurs · Three collinear points, noise 0rotation error (degrees): 9.3 -> 2.2Improved
use ours (threshold 0.05) instead of EPnP on collinear triplesOurs · Three collinear points, noise 0rotation error (degrees): 43.5 -> 2.2Improved
use ours (threshold 0.05) instead of SQPnP on planar pointsOurs · Planar configuration, noise 0rotation error (degrees): 7.2 -> 8.0No clear effect
use error-based early rejection when one 3D point is mismatchedOurs · Fast rejection experimentrejection rate: 99% rejected at threshold .05; 96% at threshold 0.1Improved
EPnP and SQPnP on mismatched points (no rejection)EPnP / SQPnP · Fast rejection experimentsuccess count: 10000 successes (EPnP), 9996 (SQPnP)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy