[FFmpeg-devel] FFV1 Specification

[Dave Rice]

On Apr 20, 2012, at 2:49 AM, Peter B. wrote:

> On 04/20/2012 01:21 AM, Michael Niedermayer wrote:
>> both ways are doable but its easy to add it to the ffv1 stream so
>> maybe we should just do that [...]
> 
> From your point of view as a digital video expert, which approach would
> you consider more reasonable?

The Federal Agencies Digitization Guidelines Initiative (a collection of federal agencies in the United States) has spent time examining approaches for encoding and storage of interlaced data within their use of lossless encoding (which is based around jpeg2000 and MXF), see page 5 here http://www.digitizationguidelines.gov/guidelines/FADGI_MXF_ASAP_Issues_20110815.pdf. As far as I know they haven't proposed a recommendation on which strategy they prefer; however their long-term practice has been to store individual interlaced fields in their own jpeg2000 frames.
Here a sample of j2k stored in this way: http://archive.org/download/sample_losslessj2k_mxf/test-jpeg2000-color-wsound1.mxf

> The great thing about FFv1 is that it's lightweight and straightforward.
> I wouldn't want to unnecessarily pack "bloat-shinystuff(tm)" into it for
> no good reason ;)

A slight advantage to store interlaced video compressed as fields rather than frames is that the video would likely compress more efficiently.
Dave Rice

> Pb
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel