COMP-8510 Visual Processing

• Setup
• Project 1
  • common.py
  • opencv.py
  • svd.py
  • both.py
• Project 2
  • pixel_matching.py
• Project 3 and 4
  • dense_pixel_matching.py
  • viewer.py

Setup

1. This will require installing Python.
   • It is recommended you create a virtual environment for Python.
2. Install the requirements.txt file for the Python packages.
   • If using pip, run pip install -r requirements.txt.
   • Optionally, manually install the required packages.
     • numpy
     • [opencv-python](https://pypi.org/project/opencv-python "OpenCV Python")
     • open3d
     • pillow
     • scipy
     • tqdm

Project 1

common.py

• This contains common helper methods for both the OpenCV and SVD Camera calibration methods.
• Running this will simply load the points from 3D.txt and 2D.txt and display them.

opencv.py

• Performs camera calibration using OpenCV for the points from 3D.txt and 2D.txt.

svd.py

• Performs camera calibration using SVD for the points from 3D.txt and 2D.txt.

both.py

• Extra helper file to run both OpenCV and SVD camera calibration for the points from 3D.txt and 2D.txt at once.

Project 2

pixel_matching.py

• Performs pixel matching by allowing you to load two images, calculating the fundamental matrix between them, and selecting a point in the left image to display the matching point in the right image.
• By default, this will try to load Left.jpg and Right.jpg from the working directory.

Project 3 and 4

dense_pixel_matching.py

• Perform dense pixel matching as done in the paper Achieving Efficient Dense Matching for Uncalibrated Images.
• Will additionally export reconsturcted 3D points.
• By default, this will try to load Simulated-Left.png and Simulated-Right.png from the working directory.

viewer.py

• View 3D points.
• By default, will try to load the ground truth values associated with Simulated-Right.png.