Add support for decoding Tiff images with associated alpha data

[brianpopow]

Prerequisites

• I have written a descriptive pull-request title
• I have verified that there are no overlapping pull-requests open
• I have verified that I am following the existing coding patterns and practice as demonstrated in the repository. These follow strict Stylecop rules 👮.
• I have provided test coverage for my change (where applicable)

Description

This PR adds support for decoding Tiff images with associated alpha data: The alpha data is pre-multiplied.

[JimBobSquarePants]

@brianpopow Do you want this in 2.1? I can hold a release 'til we get it done.

[JimBobSquarePants]

@brianpopow I updated your PR to use the built in operations. I had to tweak the tolerance slightly but the output is mostly the same (better alpha blending of edge pixels with Numerics).

I'm curious as to why there is such a significant difference from the reference as I cannot see an obvious issue.

[JimBobSquarePants]

@brianpopow I might have found quite a serious issue with the Tiff decoder I hadn't spotted before. We use FromVector4 all through the decoder codebase when we should be using FromScaledVector4. There's some cases in the Bmp decoder also.

I'm going to update and fix in this PR.

[JimBobSquarePants]

LGTM 👍

[brianpopow]

@brianpopow Do you want this in 2.1? I can hold a release 'til we get it done.

I dont think this should block 2.1 for this and it can be added for future.

@brianpopow I might have found quite a serious issue with the Tiff decoder I hadn't spotted before. We use FromVector4 all > through the decoder codebase when we should be using FromScaledVector4. There's some cases in the Bmp decoder also.
I'm going to update and fix in this PR.

Good spot 👍

Can't we use PixelOperations.Instance.FromVector4Destructive passing PixelConversionModifiers.Premultiply here? This allows us to use Avx2 unpremultiplication.

That was also my first thought, but I was getting different results then the reference decoder, so I was leaving it to the manual calculation.

Can't we use PixelOperations.Instance.FromVector4Destructive passing PixelConversionModifiers.Premultiply here? This allows us to use Avx2 unpremultiplication.

Let me double check that, but last time I checked any difference to the reference decoder was always off by exact 1, which I blamed on floating point inaccuracies.

[JimBobSquarePants]

The hardware accelerated code should be super accurate. (More so than software) and is well tested so I think the reference decoder might have an issue. That and the same method used for the other color type works just fine. (I wonder if we can compare also with System Drawing?)

I’m happy for this to go in as-is unless you can find something significant.

[JimBobSquarePants]

@brianpopow I just did a comparison using the System.Drawing reference decoder and the tests pass with the original comparison tolerance of .0.001F. This is definitely good to go.