Hi,
I would like to ask what type of majority filter (e.g. mean, median) is recommended for continuous data (land surface temperature)?
Also, does the pixel size of my raster layer (1km) affects the choice of the majority filter?
Thank you.
The pixel size does not affect the filtering locally.
If your data has speckle noise then go for median filter otherwise for smoothening, mean filter is fine.
Though the choice is highly data dependent, in my opinion, if you are dealing with very high resolution data (in range of few meters/pixel), probably median is a better choice. If you are dealing with very low resolution data (as you mentioned 1 km/pixel), probably applying a majority filter is not a good choice.
Well, I am using 8 day MODIS LST (MOD11A2) and the pixel size is almost 1km. I have a extra raster layer (let's name it r2) which I will use it as a mask for MODIS. The second layer is a binary layer (1 = urban area, null = non urban area). When I mask the MODIS raster with the r2 layer I get an image of MODIS like the one I attached. I want to clean that image from single pixels. That's my goal.
Even then, I don't think the single pixels are outliers or erroneous values. Are they?
No, they are not. Probably they are really small impervious surfaces (the above image it's a MOD11A2 product masked by the Soil Sealing imperviousness raster).
Anyway, you can use a threshold to remove those pixels. The spatial modeler of ERDAS Imagine can help with thresholding.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping