[FFmpeg-devel] [PATCH v2 1/3] libavutil: introduce AVFilmGrainParams side data

[Lynne]

Nov 23, 2020, 17:02 by dev at lynne.ee:

> Nov 23, 2020, 15:06 by jamrial at gmail.com:
>
>> On 11/23/2020 9:08 AM, Lynne wrote:
>>
>>
>>> From e11df30e25f1b27a4ec3efb7dc894b8cf59113d2 Mon Sep 17 00:00:00 2001
>>> From: Lynne <dev at lynne.ee>
>>> Date: Thu, 12 Nov 2020 12:44:30 +0100
>>> Subject: [PATCH v3 1/4] libavutil: introduce AVFilmGrainParams side data
>>>
>>> This patch introduces a new frame side data type AVFilmGrainParams for use
>>> with video codecs which are able to use it.
>>>
>>> It is generalized rather than being AV1 specific as AV2 is expected to carry
>>> the same data, as well as the fact there already exist rarely-used specifications
>>> for both H.264 and HEVC.
>>>
>>
>> This part of the commit message is no longer true, so remove it.
>>
>
> Done.
>
>
>> Looking at the HEVC spec, film_grain_full_range_flag does not have the same semantics as clip_to_restricted_range.
>>
>> If this field is going to be outside AVFilmGrain###Params, then it needs to be generic enough for it. So maybe call it color_range and make it of type AVColorRange?
>> Alternatively, just put it in AVFilmGrainAOMParams and lets not try to predict what other FG implementations are going to do in the future.
>>
>
> I've put it into AOM's struct. I think that's safer and I thought it somehow didn't
> fit in into the main one.
>
> I've added the comments you requested for both structs, fixed a few comments
> that were wrong, and as discussed on IRC changed the data types to what is
> smallest and matches best for the coefficients and generic ints for everything
> else (#coeffs, shifts, offsets and multipliers).
>
> Attached v4 and locally updated the libdav1d patch.
>

Pushed, thanks to jamrial and elenril for the reviews.