A Hough Voting based 2-Point RANSAC Solution to the Perspective-n-Point Problem

Chi Xu, Tingrui Guo, Yuan Huang, Li Cheng

Abstract

Perspective-n-point is a fundamental problem in multi-view geometry, yet two critical challenges persist: (1) The issues of high outlier rate and near degenerate cases exert a substantial impact on the robustness of existing PnP methods. In the worst-case where both issues are in presence, existing methods tend to either produce erroneous results or become computationally prohibitive. (2) Conventionally, the hypothetical pose with the maximum inlier-set is assumed to be correct. However, it remains unclear whether this assumption holds when the outlier rate approaches ultra-high levels, and along this line what is the maximum amount of outliers that can be robustly handled. To address these challenges, this paper proposes a novel Hough voting based 2-point RANSAC solution. To our knowledge, it is the first PnP solution capable of accurately and efficiently handling high outlier rates in near-degenerate cases. Extensive empirical evaluations have been conducted using the proposed approach, with a particular focus on a systematic examination under ultra-high outlier rates. The results show that, on random synthetic data, our approach works robustly even when dealing with up to 99% outliers. Meanwhile on real-world datasets, the maximum inlier-set assumption oftentimes fails when the outlier rate exceeds 97%, as the incorrect hypothetical poses may yield more inliers than the ground-truths. Our dataset and source code are to be made available at [https://github.com/xuchi7/RPnP_plusplus].

Supplementary Material

S1 S2 S3 S4

Datasets

Please download the files via the links below, then unzip the file to the root directory of this repository.

• Synthetic data SynthData.zip

• Real data RealData.zip

• Star-River Complex StarRiver.zip

Source Code

This repository contains the implementation of the proposed R2PPnP method based on Hough Voting, along with its refinement module, and the full experimental pipeline for Synthetic, Real, and StarRiver dataset experiments.

The project includes:

• Source code (src)
  • Common functions (common_funcs)
• Experimental scripts for Synthetic, Real, and StarRiver data
• All required datasets under the data/ directory (default test data and results included)

---

Directory Structure

src/                               — Source code
    common_funcs/                  — Shared functions
        r2ppnp.m                   — Proposed Hough Voting–based method
        applyFinalize.m            — Proposed refinement method
	...
   
    # files for Synth Data Experiments
    main_syn_run.m                 — Main script to run synthetic data experiments
    main_syn_show_results.m        — Display synthetic experiment results
    generate_syn_data.m            — Synthetic data generation script
   
    # files for Real Data Experiments
    main_real_run.m                — Main script to run real data experiments
    main_real_show_results.m       — Display real experiment results
    main_real_show_images.m        — Display real image samples

    # files for StarRiver Experiments
    main_starriver_show_results.m  — Show StarRiver experiment results
    main_starriver_show_images.m   — Show StarRiver image samples

data/
    RealData/
        testData/                  — Real data experiment inputs
        testResults/               — Real data experiment outputs
    SynthData/   
        testData/                  — Synthetic data experiment inputs
        testResults/               — Synthetic data experiment outputs
    StarRiver/   
        splg/   
            testData/              — StarRiver synthetic-like test data
            testResults/           — StarRiver experiment results
        images_upright/            — StarRiver upright images
    RealImages/
        brandenburg_gate/
        buckingham_palace/
        ... (totally selected 10 scenes, each containing corresponding images)

---

Running the Experiments

1. Synthetic Data Experiment

Run the experiment

main_syn_run.m

• Uses data from data/SynthData/testData/
• By default, results are NOT overwritten

To enable overwriting results:

• In common_funcs/evaluate_syn.m change line ~73:

  if 1 → if 0
  

• In common_funcs/evaluate_syn_refinement.m change line ~76:

  if 1 → if 0
  

Display results

main_syn_show_results.m

• Results are loaded from data/SynthData/testResults/

Generate new synthetic data (optional)

generate_syn_data.m

• Default data already exists in data/SynthData/testData/

• To regenerate new, change line 4:

  if 1 → if 0
  

---

2. Real Data Experiment

Run the experiment

main_real_run.m

• Uses data from data/RealData/testData/
• By default, results are NOT overwritten

To enable overwriting results:

• In common_funcs/evaluate_real.m change line ~36:

  if 1 → if 0
  

• In common_funcs/evaluate_real_refinement.m change line ~44:

  if 1 → if 0
  

Display results

main_real_show_results.m

• Experiments use data/RealData/testData/
• Results are loaded from data/RealData/testResults/

Display real images

main_real_show_images.m

• Images are located in data/RealImages/

which can be downloaded from https://www.cs.ubc.ca/~kmyi/imw2020/data.html

---

3. StarRiver Experiment

Display StarRiver results

main_starriver_show_results.m

• Inputs: data/StarRiver/splg/testData
• Results: data/StarRiver/splg/testResults

Display StarRiver images

main_starriver_show_images.m

• Uses images from: data/StarRiver/images_upright

---

Important Notes

Handling Crashes in 3rd-party Code

Some third-party code may crash and interrupt the experiment. To prevent this, restore the commented try–catch blocks in:

• common_funcs/evaluate_real.m
• common_funcs/evaluate_real_refinement.m
• common_funcs/evaluate_syn.m
• common_funcs/evaluate_syn_refinement.m

Example change:

Original (commented out):

% try
    [R1, t1, ntrial] = func(X, x, K, th_pixel, varargin{:});
% catch
%     R1 = []; t1 = []; ntrial = 0;
% end

Replace with:

try
    [R1, t1, ntrial] = func(X, x, K, th_pixel, varargin{:});
catch
    R1 = []; t1 = []; ntrial = 0;
end

This ensures experiments continue even if one method fails.

---