Add experimental gain map support

[DichenZhang1]

This feature is hidden behind the WITH_EXPERIMENTAL_GAIN_MAP compile flag.

See https://helpx.adobe.com/camera-raw/using/gain-map.html for an introduction
to gain maps. Note that in this change, gain maps are not stored as auxiliary
items as proposed by Adobe, but as a hidden input to a new 'tmap' (tone mapped)
derived item.

An 'altr' (alternatives group) box is also added to tell the decoder that it
may show the tone mapped image (preferrably) or the base image as a fall back.
The primary item id still points to the base image for backward compatibility.

[DichenZhang1]

Hi there!

Can someone help to review this pull request? This is adding gain map support for HEIC and AVIF images.

Thank you!

[DichenZhang1]

Hi there!

This change is adding support for encoding and decoding HEIC and AVIF with gain map images. Could anyone help to take a look and provide feedback?

Thank you!

[DichenZhang1]

Friendly bumping up again. Thank you!

[farindk]

Thank you for the contribution. It generally looks good.

The GainMapMetadata is currently a private data class, but also exposed in the API.
I guess that this data should also be public.
In that case, we should rename this to follow the usual naming scheme (like heif_gain_map_metadata), make it a POD struct and move it to a public header. We can add a new header file (e.g. heif_hdr.h) where we collect HDR related things so that heif.h is not getting too long.

[farindk]

Is this standardized anywhere? Could you point me to the specification?

The Adobe document says:

Gain Maps are currently undergoing standardization in ISO/TC 42 Photography. Furthermore, various
technical committees have ongoing efforts to standardize Gain Map storage in different file formats.
Consequently, the original Gain Map storage proposals have been removed from this document to avoid
confusion and conflicting information.

Does that mean that there is no final spec yet? We can include it as "experimental" anyway, but have to make it very clear that the spec is not finalized yet.

Do you have any kind of test program? Maybe a patched heif_convert that I could use to test this?

[DichenZhang1]

Hi Dirk,

Thank you for looking into this!

I've changed the GainMapMetadata name to heif_gain_map_metadata and made it public. The spec is not finalized yet, and I'll keep monitoring the updates. For verification, I've added some code in heif_enc and heif_info and they can be a closed loop for encoding and decoding. But we only have a pseudo gain map metadata (all values are zeros) and ideally this should be an input from user.

Thank you!