16-bit selection?

[RavenWorks]

I'm trying to write an image-manipulation script that involves operating several images' silhouettes on each other. My original idea was to do this mostly by moving information around using the selection buffer, and this basically worked, until it reached the point that I realised I needed 16 bits of precision for the silhouettes (there's a lot of very soft gradients, and I'm using the end result of this script in calculations, so neither banding nor dithering are acceptable). Changing my document's color depth to 16bpc wasn't producing any higher-quality results, and it took me a while to finally narrow down the reason why: even in higher color depths, Photoshop still only stores selections with 8 bits of precision!

I've found a few workarounds for this problem (mostly stuff involving locking a layer's alpha, and Apply Image, and merging layers with different blend modes) but these workarounds are all so awkward, and have fairly bad performance; does anyone know any ways to directly address the problem, or does Adobe have any plans to actually use higher precision values for selections in higher precision documents?

[D Fosse]

Yes, selections being 8-bit is a well known limitation in Photoshop. Old code no doubt, and I fully agree they should do something about this.

Masks, however, are not limited this way. They maintain the bit depth of the document. If you can work directly with masks, instead of selections, that could be a workaround.