Please give some thought to allowing a workflow to allow NOT promoting the raster bit depth even when using the "clip to polygon". Why not leave this under the user's control?
For example, we have some ordinary 8-bit BW historical aerials I want to load to SDE. I need to clip off the areas valued "0" around the edge (areas should have been "nodata") of some of the images, in order to get a better color matching result when mosaicing to SDE. (If I just load them directly to SDE using the "ignore background value" option, it works, but the imagery has bad tonal matching between photos, hence the need for the color balance/matching etc.) Note that NONE of the area to be extracted from the image is currently nodata.
If I clip using a poly and do the geometry clipping, it promotes to 16 bit, which DOUBLES the size on disk (even though NONE is currently nodata, so it should NOT need to do this). Then I'd be trying to mosaic 8 and 16 bit imagery into the same SDE dataset, so it would end up doubling the whole set to 16 bit, taking up 2x the disk space needed.
But if I set the 0 to be no data, then any black areas in the photo also get to be nodata, leaving holes.
So the only workflow I have come up with is to clip the images, then reclassify each one twice (once to set the current nodata value to 255 and the existing 255 to 254, then again to set the 255 to nodata) to force it back to 8 bit. So I waste several hours fighting this.
Why does this have to be so hard? Why can't there be an option similar to the "ignore background color" when doing the clip?