Skip to main content
Solved

RasterMosaicker Fails using Alpha Band

  • May 12, 2026
  • 6 replies
  • 89 views

mifranz
Contributor
Forum|alt.badge.img+4

I have a workbench with a RasterMosaicker that runs fine in 2021.2.6, but the same workbench throws an error in 2025.1 “Failed to mosaic the rasters”. Same error for version 2024.1.2.1.

The error occurs only when setting the parameter “Overlapping Values” to “Composite Using Alpha Band”.

Is there anything known causing issues regarding raster transformers in newer versions that might explain this?

Best answer by mifranz

I have reordered the bands using RasterBandReorderer from ARGB to RGBA, now it has worked. I’m not yet sure if that has any other undesired consequences for my process, but maybe the RasterMosaicker got stricter in terms of band order and the Alpha band is supposed to be in last position while it previously did not matter?

6 replies

j.botterill
Influencer
Forum|alt.badge.img+58
  • Influencer
  • May 12, 2026

Can you inspect the feature/record going into the RasterMosaicker and take a screen snapshot of the FeatureInformation showing the raster and its Alpha band

You need to ensure all features/records have this alpha band

FME2024.1 might be stricker about mixed bands being tagged or not tagged with Alpha

Use RasterSelect to ensure band count is the same… forcing all RGB to RGBA

or RasterBandNodataSetter (for nodata → alpha effect)


mifranz
Contributor
Forum|alt.badge.img+4
  • Author
  • Contributor
  • May 12, 2026

Here are the screenshots of the details of the two input rasters, each with an Alpha8 and RED8, GREEN8, BLUE8 band. 
I’ve also tried the RasterBandNodataSetter as well as the RasterBandNodataRemover, but the Mosaicker fails regardless. None of the bands have a palette.


j.botterill
Influencer
Forum|alt.badge.img+58
  • Influencer
  • May 15, 2026

Well, they both have the same coordinate system, extents, cell size and origin.

FME must be misinterpreting the alpha band

Prior to the RasterMocaiker, can you inspect both have the same numbers

  • fme_raster_numbands
  • fme_raster_band_interpretation

mifranz
Contributor
Forum|alt.badge.img+4
  • Author
  • Contributor
  • May 15, 2026

Both rasters seem to have the same values.

Only thing that seems to be different is the attribute band_num_tile_rows / band_num_tile_columns which is 1024 for the alpha band and 8192 for the RGB bands. But in the listed resolution of the whole raster it is 8192. And when extending the info about the raster, all four bands are listed with 8192 so idk if this is actually an issue.


mifranz
Contributor
Forum|alt.badge.img+4
  • Author
  • Contributor
  • Best Answer
  • May 15, 2026

I have reordered the bands using RasterBandReorderer from ARGB to RGBA, now it has worked. I’m not yet sure if that has any other undesired consequences for my process, but maybe the RasterMosaicker got stricter in terms of band order and the Alpha band is supposed to be in last position while it previously did not matter?


kailinatsafe
Safer
Forum|alt.badge.img+23

Hello ​@mifranz, good catch! There indeed was a behaviour change in FME Form 2022.1 that tightened up validation in the RasterMosaicker so that mismatched interpretations will cause a failure or be rejected, whereas older versions were more lenient. I will consider filing an issue to have the documentation updated. The RasterBandOrderer is likely the best bet before sending the rasters to the mosaicker. Happy to help, Kailin.