https://github.com/arthurdouillard/teledetection

Implementation in C of a custom k-means for clouds detection in satellite images.
https://github.com/arthurdouillard/teledetection

c image-processing kmeans-clustering satellite-images

Last synced: 2 months ago
JSON representation

Implementation in C of a custom k-means for clouds detection in satellite images.

• Host: GitHub
• URL: https://github.com/arthurdouillard/teledetection
• Owner: arthurdouillard
• Created: 2017-12-13T09:05:41.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2017-12-17T16:31:25.000Z (about 7 years ago)
• Last Synced: 2024-10-12T04:29:12.397Z (3 months ago)
• Topics: c, image-processing, kmeans-clustering, satellite-images
• Language: C
• Homepage: http://adsabs.harvard.edu/abs/2001SPIE.4168...67B
• Size: 9.38 MB
• Stars: 4
• Watchers: 2
• Forks: 3
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Cloud detection with custom k-means

Implementation in C of the paper

[Introducing spatial information in k-means algorithm for clouds detection in optical satellite images](http://adsabs.harvard.edu/abs/2001SPIE.4168...67B).

A copy of the pdf is included at the root of the repo.

## The algorithm

In a few words, we need a way to compute the % of clouds in an photo taken from space.

The algorithm creates blob of pixels (central, up, right, down, left) and applies

a k-medians on it.

The k-medians is like a k-means but uses a median operation instead of a mean

operation. Furthermore the distance metric used is a Manhattan distance.

The pixels belongings to the highest k-medians' center (thus having pixels with

high values -> whiter) are considered as clouds.

## Results

![Example 1](examples/ex1.png)

Percent of clouds: 1.19 %

![Example 2](examples/ex2.png)

Percent of clouds: 5.28 %