[FFmpeg-user] When to determine frames are progressive or interlaced ?

Tim Nicholson nichot20 at yahoo.com
Tue Jan 15 08:38:44 CET 2013

On 14/01/13 10:23, Carl Eugen Hoyos wrote:
> Christian Ebert <blacktrash <at> gmx.net> writes:
>>>>> I suspect (and it is what James writes above) that mediainfo 
>>>>> only tells you if the video was encoded using an encoder 
>>>>> setting "interlaced" or not. It does not tell anything about 
>>>>> the actual content.
>> Right. mediainfo only inspects metadata, I guess.
> "metadata" is not right word imo.
> When you encode mpeg2video, you have to decide if you want to 
> encode "progressive" or "interlaced" - this has not necessarily 
> anything to do with the content, but with the used algorithms.
> Using the right algorithm increases encoding efficiency, using 
> "interlaced" when encoding progressive material (this is what 
> all television providers do afaict) is not very expensive, 
> encoding interlaced input without specifying it to the encoder 
> can be expensive but is of course possible without visible 
> quality loss using a small enough quantiser.

Although coding "progressive" as interlaced is "not very expensive" it
tends to be the safer option, as the opposite way round can cause issues
if, as you say, the quantiser isn't small enough.

However surely the bigger issue is with scaling?

Again using interlaced scaling is the safer option, but used on
progressive material can severely impair quality.

On a frame with little movement, say a news studio with a talking head,
there is much to be said for adaptive scaling/coding. Those parts of the
picture with no movement can be handled as progressive, those with
movement as interlaced.

This is how the BBC R&D 3 field adaptive scaler works, and it is
regarded as producing the best quality results. It would be good if mdsh
could his port working....

> [...]


More information about the ffmpeg-user mailing list