+13
I have a set of overlapping rasters that need to be combined, normally its not a problem, but these rasters overlap in such a way that no sorting can help.
I was wondering if someone knows of a easier solution, before diving into the rasterexpressionevaluator.
Cheers,
Itay
+13
+2
http://fmepedia.safe.com/articles/Samples_and_Demos/Alpha-Compositing-Blending-Two-Raster-Images
+13
Thanks for replaying. The option (Nodata Overwites Data) on the RasterMosaicker does not solve my issue.
Firstly the nodata value is returned as N/A when quering the rasters, setting the nodata to one value (255) does not solve the issue since there are also values rangng from 245 to 255 in the nodata region.
I did manage to mosaic them correctly by changing the range of values into nodata, but also the data region is affected (undesired).
What I need is to create a mask where the nodata can be set.
I have had a look at this example which is not very clear on how to create such a mask.
Any help would be appreciated.
+17
+13
Repalcing a range of values (240 to 255) to 255 and setting 255 to nodata, does not help since the data region also gets affected, which is undesired.
Thats why creating a mask where the values convertion can take place might be a solution, however I am still not managing to create a mask for the region 'nodata' region in the overlap.
Itay
+13
Hopefully my computer can take it, some twiking will probably help
+25
I'm assuming then, that each raster has a border around it, with values set to 255, but you can't just set that to NODATA because 255 is also used in the data itself?
If so, perhaps try set NODATA to 255 and then use a RasterExtentsCoercer to extract the Data Extents. That would give you a boundary with which to clip the black border off the raster.
If there are NODATA values in the main data area I imagine (though I haven't tried this) that they would appear as donuts or holes in the boundary. Just drop the holes and it will ensure those cells are kept when clipped. You then just need to un-set the NODATA value.
I hope that I've interpreted your scenario correctly and that this helps.
Mark
+13
Unfortunatly the border values are not only 255 and that is where mosaicking them goes wrong.
The majority of the border area contain 255 values, but closer to the data area a range of values (240-254) can be found, I imagine that it is due to the mosaicking and resampling of the original (smaller) mosaic's I am using.
I will give the RasterExtentsCoercer a shot with nodata set to 255 and see if that can be used or needs to be corrected (generalized/drop holes).
Cheers,
Itay
+13
The following method to create the data extent works:
Thanks all for the tips and guidance.
Itay
+5
In the past I have found the only way to set no data in low bit depth rasters is adjust one of the pixel values to free up one of the 255. For instance replace all '0' with 1 and then set your no data as 0. The difference between 0 and 1 is indistinguishable. If you set no data as 0 without adjusting the true '0' pixels first you will end up with transparent parts to the image in GIS software.
+16
Itay, if you have polygons for the regions where the raster should contain data, you could use the Clipper to clip the raster before mosaicking.
What is the simplest way to create an overlapping mosiac using fme? I have two images which overlap and also have no data values appearing as black on the outside. I need to remove both so the image blends together. I wasnt sure if the alpha band would be the best way to go?
+13
@diggs12 - I'd suggest checking out https://knowledge.safe.com/articles/1247/raster-mosaicking-scenario.html, and then the suggested related articles listed on the right. Definitely this is bread and butter FME, you just have to work through your no-data and alpha band situation.
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.
