For detect and removing cloud for image, there are many methods ranges from simple and complex. The simplest methods is the closest spectral fit (CSF) technique that can replace clouds and cloud shadows (CCS) pixels with the most similar pixels at cloud-free areas in the same image. The CSF technique can be applied to remove CCS pixels in the image, and then error diagnostics is conducted. On the other hand, many researchers have used empirical algorithm for removing cloud, as the following papers:
Thanks for your message and for providing article related to Ackerman algorithm. Actually at this time, I am looking for easiest way to remove clouds, and if I did not find then I have to apply brightness temperature tests to detect clouds. BTW, MOD35 is also based on Ackerman et al. but I am little confused how to highlight / detect clouds using it.
Dear Said,
Thanks for your input here. I have already tested the first article (IEEE) but the mentioned thresholds are not working for my area. I worked for few thresholds and they are working well for thick clouds but unable to work when fine aerosols mixed up with thin layers of clouds. And, these thresholds are not working properly for other images due to change in the atmospheric conditions as well as viewing geometry. I am looking for already developed thresholds so that I can easily apply over my region or someone guide me how to use MOD06 or MOD35.
I reckon the best and safest way is to use MOD35 if you're going to use MODIS images.All other method are somehow ambiguous and not necessarily always easy to use.
Could you please guide me how to use MOD35 to remove clouds from MODIS images. I found that cloud mask has five bands but I am confused to select suitable band.
MOD35 only gives you the percentage of image cloud contamination.Then if you look for clear sky you should use 100 clear condition although there might still some patches of cloud present in the pixel that brings some uncertainties to your results. To get rid of these uncertainties you need to use a LUT of the region of your interest plus a combination of optical and thermal channels. To be frank, there is no definite algorithm for this task and all depends on the skill of the person in charge all because of the ambiguous and unpredictable behavior of the clouds with the EM waves. However if you do not need too much precision,then MOD35 would suffice.
I am currently in the process of developing a new very easy method to remove clouds from a L1B scene. If you could provide me with a link of the your L1B file that you wish me to test my algorithm on so that we can see whether the desired outcome has been reached for both of us. I will then send you a jpg image of the result to judge its quality. Make sure the L1B incl;includes the following bands: