[FFmpeg-devel] [PATCH] ffplay: support rotated video.

[Clément Bœsch]

On Fri, May 23, 2014 at 12:52:42AM -0400, Don Moir wrote:
> 
> ----- Original Message ----- From: "Don Moir" <donmoir at comcast.net>
> To: "FFmpeg development discussions and patches" <ffmpeg-devel at ffmpeg.org>
> Sent: Friday, May 23, 2014 12:40 AM
> Subject: Re: [FFmpeg-devel] [PATCH] ffplay: support rotated video.
> 
> 
> On Thu, May 22, 2014 at 09:30:05PM +0200, Clément B?sch wrote:
> >On Wed, May 21, 2014 at 05:18:30AM +0200, Michael Niedermayer wrote:
> >> On Tue, May 20, 2014 at 07:54:05PM +0200, Clément B?sch wrote:
> >> > On Fri, Apr 18, 2014 at 05:43:35PM -0400, Don Moir wrote:
> >> > [...]
> >> > > >+    if (autorotate) {
> >> > > >+        AVDictionaryEntry *rotate_tag = av_dict_get(is->video_st->metadata, "rotate", NULL, 0);
> >> > > >+        if (rotate_tag) {
> >> > > >+            if (!strcmp(rotate_tag->value, "90")) {
> >> > > >+                INSERT_FILT("transpose", "1");
> >> > > >+            } else if (!strcmp(rotate_tag->value, "180")) {
> >> > > >+                INSERT_FILT("hflip", NULL);
> >> > > >+                INSERT_FILT("vflip", NULL);
> >> > > >+            } else if (!strcmp(rotate_tag->value, "270")) {
> >> > > >+                INSERT_FILT("transpose", "2");
> >> > > >+            } else {
> >> > > >+                INSERT_FILT("rotate", rotate_tag->value);
> >> > > >+            }
> >> > > >+        }
> >> > > >+    }
> >> > >
> >> > > you might want to do this instead of stcmp. Catches values greater than 360 as well.
> >> > >
> >> > > int rotate = atoi(rotate_tag->value) % 360;
> >> > >
> >> >
> >> > No I can't, it can be a floating value.
> >>
> >> fmod() could be used
> >>
> >
> >>Right, but I'm not sure it would be more reliable/simple that what i'm
> >>currently proposing.
> 
> >Not sure if it matters if its floating or not.
> 
> >atoi ("180.0") equates to 180 if that is what you are refering to.
> 
> >With what you have a filter gets inserted for anything your logic does not
> >understand including zero rotate. doing a mod on it catches zero easily
> >and anything >= 360. Probably not a big deal except the inserting of
> >filter for possible zero rotate value.
> 

> I guess what you are saying it might be something like "90.5" or something.

Yes. 90.999 would raise 90 too unless you round btw.

> Could be one degree off. Not sure it would display where someone would
> notice a one degree difference.

There is no reason to drop that accuracy. The orientation matrix in MOV
has a relatively high one, there is no reason to drop it just to save 3
strcmp on 2-3 characters called only once at the beginning of the program.

> Can it be radians and how would you know
> that?
> 

See the other thread where I sent the updated patchset. It should answer
all your questions.

> Do you have a video with an odd rotate value ?

Not yet, but it's almost free to deal with that. 90, 180, and 270 cases
are just exception, and I don't want to catch them because of an
inaccuracy. It's more complex to write a matrix leading to 90.1 degrees
than 90 degrees, so there is no reason to rotate 90 instead of 90.1 in
that situation.

-- 
Clément B.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20140523/26e12b56/attachment.asc>