On 14/04/14 08:18, Niels Möller wrote:
> Vittorio Giovara <[email protected]> writes:
> 
>> +    /**
>> +     * This side data contains a 3x2 matrix (math notation) describing
>> +     * two dimensional transformations to be applied on the decoded video.
>> +     * Elements are represented as s32le, fixed point 16.16.
>> +     */
>> +    AV_PKT_DATA_DISPLAYMATRIX,
> 
> It's 3x2 because you also include translation? Fixed-point 16.16 sounds
> like a waste for the matrix elements for rotations (elements should be
> in the range -1/+1). Or do you want to support some scaling as well
> here?

Apparently so.

> Since useful range and precision are quite different for rotation data
> and translation, maybe it would make sense to separate them?

16bit of precision is already a lot for the current use-case.

lu

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to