Panoramic Construction with SIFT

1001532 - Computer Vision (2022/1)

DC - Department of Computing

UFSCar - Universidade Federal de São Carlos

🔵 Introduction

Practical activity proposed during the course of Computer Vision (1001532), offered by the Department of Computing at UFSCar in semester 2022/1.

This project stitches images together to create a Panorama. Images are taken from different perspectives and transformed to fit together by using SIFT feature detection and RANSAC algorithm for homography transformations

🔵 Requirements

Before you start, make sure you've met the following requirements:

pip install -r requirements.txt

🔵 Installing Panoramic Construction with SIFT

• Install requirements
• Import panorama.py
• The folder pano show the results

🔵 Using Panoramic Construction with SIFT on terminal

• See example.py for how use the module

Panoramic of image set 1

Panoramic of image set 2

Panoramic of image set 3

Panoramic of image set 4

🔵 Using Panoramic Construction with SIFT on notebook

• Apresentação.ipynb execute the project on python notebook
• Teste_Ruído.ipynb executes noise test on the implementation
• Sintônia_de_Parâmetros.ipynb executes some parameter tunning and analyses

🔵 Authors

Heitor Souza

Gustavo Chavari

🔵 License

This project is under license. Acess the file LICENSE for more details.

🔵 References

• Introduction to SIFT (Scale-Invariant Feature Transform)

• 📄 LOWE, David G. Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision, v. 60, n. 2, p. 91–110, 2004.

• 📄 REY-OTERO, Ives and DELBRACIO, Mauricio. Anatomy of the SIFT Method. Image Processing On Line, 2014.

• 📽️ SIFT Detector | SIFT Detector

• OpenCV: Feature Matching + Homography to find Objects

• 📄 D. Li, D. Han, X. Zhang and L. Zhang, Panoramic image mosaic technology based on SIFT algorithm in power monitoring, 2019 6th International Conference on Systems and Informatics (ICSAI), 2019, pp. 1329-1333, doi: 10.1109/ICSAI48974.2019.9010294.

• stackoverflow: how to show the whole image when using opencv warpperspective