Plane Sweeping 2-View Stereo

Prerequisites

• Python 3 (version: 3.7.1)
• Python packages: cv2 (version: 3.4.3), numpy, os, glob
• MAC OS or Windows 10 (Ubuntu Windows' Bash)

Installation

• Clone the repo:

git clone https://github.com/msouppe/Two_View_Stereo.git

• Navigate to main.py

• Run the program:

python3 main.py 

Project Structure

• main.py: Main program
• calibration.py: Calibrates camera using the chessboard images
• camera_pose.py: Craw eplipole lines on images, re-projected points on images, and plane-sweeping stereo computations
• images: All images that were used for the project i.e) chessboard images, scene images, warped images, undistorted images

Output

Part 1
For the images that were taken to calibrate the camera we obtain the following:

• Intrinsic matrix K
• Radial distortion coefficients
• Reprojection mean square error

Part 2
None, only needed to take photos of a scene

Part 3
Computation for finding the camera relative pose:

• Few epipolar lines on the images.
• The matrix R^R_L and r^R
• Re-projected feature points on the first image

Part 4
Creating a plane-sweeping stereo

• Values for d_min and d_max
• N=20 warped second images
• Resulting depth image in grayscale