On Thu, Feb 16, 2017 at 11:09 AM Vittorio Giovara < vittorio.giov...@gmail.com> wrote:
> On Wed, Feb 15, 2017 at 2:54 PM, Aaron Colwell <acolw...@google.com> > wrote: > > Hi Vittorio, > > > > This may not be a blocker for this patch, but one issue with converting > the > > bounds to pixels like you do here is that resizing a video could result > in > > incorrect metadata being generated when muxing. If you keep the bounds in > > the 0.0-1.0 fixed point space this problem doesn't happen since it is a > > resolution independent representation. If you still want to use pixels > here > > then the resizing filter will need to become aware of AVSphericalMapping > and > > adjust it accordingly. I think cubemap padding in pixels may have a > similar > > issue. This is no way intended to be a blocking comment. It is just > meant to > > raise awareness on something you might not have considered. > > Hi Aaron, > this is an interesting point, but leaves up open questions. Do you > foresee actual cases in which this is going to be a problem? It's not > the first time we have fixed point fields exported, but it's also true > that the conversion from a 0.32 system is not very common and could > leave users puzzled. Regarding padding, is the specification going to > be updated to fix this potential issue? > Thank you > -- > Vittorio > Hi Vittorio, After sleeping on this, I think what you have will be fine. Resizing a cubemap w/ padding is just going to have to be handled in a special way because fractional pixel padding isn't going to work right on the GPU. You really only want to waste a few pixels on padding, not a constant fraction of the whole frame. Given that we have to handle cubemaps in a special way for resizing, then my thoughts about resizing equirect aren't really a big deal. I don't expect any spec changes will be needed for this. I suppose this just a long way of saying "we will still need a followup for resizing spherical videos properly, but this patch LGTM." Thanks again for working on this. Aaron _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
