Implementation of the 8-point algorithm to estimate relative camera pose (rotation and translation) from image pairs using fundamental and essential matrix decomposition.

This project implements multiple methods for camera pose estimation from a sequence of images:

• 8-Point Algorithm (least squares)
• RANSAC-based fundamental matrix estimation
• OpenCV implementation for comparison
• Camera trajectory reconstruction (iterative and non-iterative)

• Feature detection and matching using SIFT/ORB
• Fundamental matrix computation (normalized 8-point, RANSAC)
• Essential matrix recovery from intrinsic calibration
• Rotation and translation disambiguation using triangulation
• 3D camera trajectory visualization

Image before camera translation:

Image after camera translation:

Corresponding features detected and matched between the two images using SIFT descriptors:

Core Components:

• EightPoint.py - Main implementation with 4 methods:
  • getRotationTranslationFromImagesRANSAC() - RANSAC-based fundamental matrix
  • getRotationTranslationFromImagesLS() - Least squares fundamental matrix
  • getRotationTranslationFromImages8Point() - Classic 8-point algorithm
  • getRotationTranslationFromImagesOpenCV() - OpenCV's implementation
• processSequence.py - Batch processing for image sequences
• test.py - Example usage and visualization

Methods:

1. Extract keypoints and descriptors (SIFT/ORB)
2. Match features between image pairs
3. Compute fundamental matrix F (8-point or RANSAC)
4. Recover essential matrix E = K^T F K
5. Decompose E into 4 candidate (R, t) solutions
6. Disambiguate using point triangulation (cheirality check)

from EightPoint import EightPoint
import cv2
import numpy as np

# Initialize
eight_point = EightPoint()

# Load images
img1 = cv2.cvtColor(cv2.imread('image1.png'), cv2.COLOR_BGR2GRAY)
img2 = cv2.cvtColor(cv2.imread('image2.png'), cv2.COLOR_BGR2GRAY)

# Camera intrinsic matrix
K = np.array([[focal_x, 0, cx],
              [0, focal_y, cy],
              [0, 0, 1]])

# Estimate pose
R, t, scale = eight_point.getRotationTranslationFromImagesRANSAC(img1, img2, K)

See requirements.txt for the full list of dependencies. The requirements file is auto-generated and contains the core packages used by this project.

Install dependencies with:

pip install -r requirements.txt

The implementation was tested on synthetic rendered sequences with known ground truth. Results show:

• RANSAC provides robust estimation in the presence of outliers
• Iterative pose estimation accumulates drift over long sequences
• Direct estimation relative to first frame reduces accumulation error

See /Test1/ directory for detailed CSV results comparing different methods.