from .mxf HD to .mxf SD

[marcorocchini]

Hi

considering my usual https://dl.dropboxusercontent.com/u/39871584/C0020.mxf

can I use ffmpeg to pass from HD (of course 16:9) to SD 16:9 and 4:3 (sd mxf= IMX D10 format of ffmbc)?

[poisondeathray]

Do you remember this thread?
https://forum.videohelp.com/threads/362968-ffmpeg-from-mp4-to-mxf

It's basically the same thing, piping ffmpeg to ffmbc

In ffmpeg, you need to deinterlace (double rate to 1080p50), progressive scale to SD, change to SD color (BT601), re-interlace; then that is piped to ffmbc to encode IMX50

ffmpeg -i input.mxf -vf yadif=1:0,scale=720:576,colormatrix=bt709:bt601,interlace -pix_fmt yuv422p -an -f rawvideo - | ffmbc -f rawvideo -pix_fmt yuv422p -s 720x576 -r 25 -i - -tff -target imx50 -aspect 16:9 output.mxf

You can do other processing as well e.g. you might want to lowpass for SD interlace etc..., to reduce line twittering or shimmering , and/or use a softer resizer

There is a new ffmbc out a few days ago synced to a more recent ffmpeg snapshot, but it needs to be compiled for windows. It might have SMPTE metadata timecode (ISMT) handling like current ffmpeg, and might have -vf interlace . If so, then you wouldn't need ffmpeg to pipe . Or newer ffmpeg builds might incorporate IMX preset in the future (then you wouldnt need ffmbc)

[racer-x]

I think he wanted 4:3 AR, so it will need to be cropped.

Out of curiousity, try this out:

[poisondeathray]

The 4:3 crop case is almost the same thing , just crop it to 1440x1080 before the resize , and change the -aspect 4:3 for the IMX50 encode. There are no special rules for horizontal cropping of interlaced material, so you can do it before the deinterlace (fewer pixels to process, thus faster)

ffmpeg -i input.mxf -vf crop=1440:1080:240:0,yadif=1:0,scale=720:576,colormatrix=bt709:bt601,interlace -pix_fmt yuv422p -an -f rawvideo - | ffmbc -f rawvideo -pix_fmt yuv422p -s 720x576 -r 25 -i - -tff -target imx50 -aspect 4:3 output.mxf

[marcorocchini]

There is a new ffmbc out a few days ago

mm thank you, where is can downloadable?

[marcorocchini]

but it needs to be compiled for windows.

ahhn

[marcorocchini]

please I wonder: during HD-->SD conversion I loss quality? Or for better explain: if something needs to be broadcast in SD, should shoot directly in SD or HDtoSD conversion will produce SD worse results than shooting directly in SD

[poisondeathray]

please I wonder: during HD-->SD conversion I loss quality? Or for better explain: if something needs to be broadcast in SD, should shoot directly in SD or HDtoSD conversion will produce SD worse results than shooting directly in SD

It depends on many factors including, the camera, recording format. Some cameras do a poor job internally of HD=>SD. Some acquire at SD by subsampling the sensor, skipping lines/pixel binning, others do a proper resize

Almost always, a properly done progressive HD=>SD will yield better results, and gives you more options in post production and delivery. Interlaced HD=>SD conversions can be more tricky, there are other issues like aliasing, twittering . You can get worse results if not done properly

As always, you have to test your workflow and use what you think is better