Skip to content
/ deblur Public

Motion deblurring from a single image: blind and non-blind deconvolution

Notifications You must be signed in to change notification settings

gpl27/deblur

Repository files navigation

Implementation of High-Quality Motion Deblurring from a Single Image (SIGGRAPH 2008)

This repository refers to a project for the course INF01046 - Image Processing Fundamentals at INF in UFRGS.

This is link to the Original Paper

Brief Documentation

  • blind/ - contains some images generated from our blind deconvolution
  • non-blind/ - contains some images generated from our non-blind deconvolution
  • examples/ - contains a few examples of blurred images and their respective blurred kernels
  • fails/ - contains some fails generated during work
  • convolve.py - contains the methods related to convolution
  • deblur.py - contains the methods to perform the deconvolution
  • helpers.py - contains some methods for reading images and kernels
  • main.py - performs the algorithm

For a more detailed and comprehensive explanation of this project, please refer to this document.

Results

Non-Blind Deconvolution

These are the results of applying non-blind deconvolution (the input is both the blurred image and the kernel) to certain images:

Original Image Blurred Image and Kernel Our Results
Original Image Blurred Image and Kernel Our Results
Original Image Blurred Image and Kernel Our Results
Original Image Blurred Image and Kernel Our Results

Blind Deconvolution

These are the results of applying blind deconvolution (the input is only the blurred image) comparing the results with the original paper:

Blurred Image Our Results Original Paper
Blurred Image Our Results Original Paper
Blurred Image Our Results Original Paper
Blurred Image Our Results Original Paper
Blurred Image Our Results Original Paper

About

Motion deblurring from a single image: blind and non-blind deconvolution

Topics

Resources

Stars

Watchers

Forks

Languages