Median filter is used (commonly )

https://www.rdocumentation.org/packages/fractal/versions/2.0-4/topics/medianFilter

Median filter is used (commonly )

https://www.rdocumentation.org/packages/fractal/versions/2.0-4/topics/medianFilter

you can remove it by using some types of enhancement filter such as min, max, median, adaptive median filter. adaptive median filter is the most recommended filter.

In Matlab, you can use a binary version of your image and use bwareaopen to remove objects smaller than the desired area. You can later use this binary image as a mask for your original image. I believe R has a similar function computeFeatures.shape() that gives you various shape parameters like area that can be used to filter out the unconnected pixels.

Thanks a lot for your help. I will try out and see if your suggestions work and give you an update. It might be that my data is a bit different because I am only having single unique classes and no floats, but I will see if I can apply your advice. Unfortunately I do not have any access to Matlab... But thanks also for that tip.

Dear lisa

wavelet families (e.g. db,Bior,Harr) and median filters are suitable for this purpose.

Thanks a lot for your help Aparna Sathya Murthy, Hiba Basim Alwan Hussain Al-Dulaimi, Ruchit G Shah and Mohammad Mahdi Feraidooni. I think your solutions are actually really good. But for my data unfortunately a bit too coarse. I use the clump command in R raster package now, which worked pretty fine and can be used just for single classes. Thanks again for your inspiration!