landsat8 cloud/shadow detection

Hi everybody ,

I am working on my master's thesis entitled: cloud and cloud shadow detection in Landsat8 images, but I'm still at the stage of choosing the method to follow, I rather opt to use Matlab as a language for my program, but I do not know how to proceed.

if you can give me an idea about simple methods to implement (that I do not have to use the: multi-temporal, which means that I do not have images of different dates for the same scene).

here is an example of an image used

thank you so much

Answer 1

[Image Analyst]

To review prior algorithms, see Vision Bibliography

Search it and find published papers. Of course, you'll have to invent a new one or improve on those for your masters, not merely implement a published algorithm. What new ideas do you have so far?

[Zakaria Sadok]

hello , actually i still have one year and half , and my supervisor told me to check the existed methods and try at least to choose which algorithm i will work with and try to implement it first then i can work on improvement that is not really necessary it depends on the time , so i find myself a little lost with existant methods and the one that we can easily implement with matlab

thank you for your answer

[Image Analyst]

I don't know what the published algorithms do, but I'd use all the spectral images you can and then get a spectral image, where each pixel is the spectrum of that part of the scene. Then get the spectrum of known cloud pixels. Plot them all and see. Maybe you can get it down to just a few spectra by averaging some of them, like a white cloud spectrum and a gray cloud spectra. Then you could compare the spectra of a pixel and see how much it's like the cloud spectra. Not sure if the spectra changes scale or moves up and down (brighter and darker) so you may have to normalize the spectra to do an intensity-independent comparison. Then compare them, like with immse() or something more sophisticated, to see if that pixel is a cloud pixel or not.