ECE 285: Computational Imaging Systems - Final Project

Turning Back Time: Computational Techniques for Star Trail Removal ​from Long-Exposure Astrophotography ​

Deconvolution Results. A - Original Image, B - Wiener Deconvolution, C - Richardson-Lucy Deconvolution, D - Vanilla Gradient Descent, E - FISTA, F - FISTA + TV, G - ADMM, H - ADMM + TV

Team Members

• Nikhil Gandudi Suresh
• Avanti Bhandarkar

Documents

Presentation Slides
Project Proposal
Project Report

File Structure

• Data stores all the images we used for this project (real and synthetic).

  • binary_timed - clustered version of timed
  • real - unused images (except star3.jpg)
  • synthetic - randomly generated pixel image
  • timed - all images with time of exposure metadata available

• Documents - stores all our submitted documents

  • Proposal
  • Report - TO ADD
  • Slides

• Notebooks - stores our Jupyter Notebooks

  • image_recovery - recover the real image with star trails removed + replaced with reconstruction
  • main - contains all the utils functions, preprocessing pipeline, deconvolution manager and some outputs (real and synthetic)
  • star_clustering - K Means binary clustering and user center selection code