Support for 15-bit and 16-bit TGA file formats

[HarryLovesCode]

Hello,
I noticed recently that the TGA decoder didn't support 15-bit and 16-bit formats. This led me to ask the question, does this crate handle non-8-bit components? From what I understand, it doesn't (see dynimage.rs).

So, if I were to implement support for TGA file formats that use 5-bits per component, how would I need to represent them? Should we add support in the DynamicImage file I linked, or should they be scaled and cast to u8s?

[nwin]

There is limited support for 16-bit formats. DecodingResult::U16 exists but it is not used in dynimage.rs#L450. The most sensible approach would be to just scale it down. :)

When implementing reading uN, (N%8 != 0) image types you have two options. Scale it to the next bigger uint and lie about the ColorType. Or just read it into an u8 and modify dynimage.rs#L450 in order to scale it, which would be the better approach. Have a look at dynimage.rs#L473 on how 1, 2 and 4 bit grayscale images are handled.

Contribution are very welcome. :)

[philipc]

The BMP decoder is currently scaling to u8 and lying about the ColorType. It can handle any non-8-bit components. (It's also scaling down in some cases which isn't ideal, but those are unusual BMP formats.) @nwin Is it worth moving that scaling out of the BMP decoder? My main concern with that is it might be slower, but not sure how much that matters.

[nwin]

You’re right it’s not worth it.