Basic change detection in Remote Sensing

In this video I show a basic change detection scheme for Remote Sensing images.
I show an example with 2 CBERS-4/PAN5 images from the same place, one from 2015 and other from 2018, calling them It1 and It2.
I propose an adaptation of NDVI formula and call it NDTS (Normalized Difference of Time Series) which is basically:
NDTS = (It2 - It1)/(It2 + It1)
Then we square the result and apply a visual threshold followed by mode filtering to highlight the detected change between the two images.

The LaTeX equations and source code for algorithms of this presentation are available at:

The presentation is available at:

The two images are available at:

[XinhLe]

so glad that I found this. Thanks so much!

[SH-gzxk]

thanks a lot! question: how do you do the filtering?

[Funnyshoes]

Thanks for the video! I didn't entirely understand from the video - what is the theory and motivation behind squaring NDTS?

[rohitsharma-iesz]

I have two images of 1.7gb each and when I perform image differencing using pillow it's say out of pixel range.. what's the solution to that? And will this method work for such large images of Tiff format

[GhadeerMohed]

Hello .. I want to use this formula (ndts) in my work have you published an artical using it to cite it as one of my refernces??

[dhwanitrivedi]

Thank you so much for explaining. How to do the same process using Machine Learning?

[zahirazahora]

how i hould do the filtering mode in order to avoid the noie