[FFmpeg-user] Understanding ffprobe output

Phil Rhodes phil_rhodes at rocketmail.com
Sat Nov 28 18:30:20 EET 2020


 

    On Saturday, 28 November 2020, 15:49:09 GMT, Simon Roberts <simon at dancingcloudservices.com> wrote:  
> So... the term "field" isn't used to describe one half of an interlaced> frame? 'coz I have devices that claim to output progressive and interlaced> as a choice, and they're not CRTs..
This is a somewhat vexed question that, in the end, comes down to semantics.
For the sake of providing a comprehensive overview, let's take it from the very start, which most people will already know but let's be complete.
In the beginning, analogue video was transmitted wirelessly and displayed on cathode ray tubes. Frame rates were 25 to 30 frames per second. At that rate, the image would have flickered visibly, but increasing the frame rate to compensate would require faster electronics and more radio bandwidth, which was considered undesirable. As an alternative, the decision was made to first send all the odd lines, and then send all the even lines, in two separate fields. The result is sometimes visible as a slight vertical shimmer in interlaced video displayed on a CRT, more so at 25fps than 30fps. Crucially, it does mean that odd lines and even lines are not photographed by a camera at the same instant in time, so the resulting pictures have some of the motion rendering characteristics of a 50fps video.
The first complexity in this matter involved transferring film to video. Film can be shot at 50fps and each frame transferred to a single video field, but this is rarely done. It's much more usual to shoot film at 24fps (or 25fps in regions traditionally using 25fps video) and use the same film frame for both fields. Traditionally, on a CRT, this means there is still interlace shimmer but that both fields were photographed at the same time. This produces altered motion rendering which is a large part of the "film look." The practice of transferring 24fps film to 30fps formats uses a technique called 3:2 pulldown which is outside the scope of this discussion, although it is possible to shoot film at 30fps and end up with the same situation, technically.
When companies such as Sony started making video equipment aimed at replacing film, capable of shooting at 24fps and without interlacing, lots of equipment still expected to receive all of the odd lines first, then all of the even lines afterward. Particularly, CRT monitors were still around, and would have suffered the old flicker problem if they'd tried to display 24fps images without interlacing. To compensate, the simple decision was made to interlace the image, sending the odd lines then the even lines, even though the camera had photographed them all simultaneously. Sony called it "progressive segmented frame," PSF.
This is convenient because a digital video stream containing interlaced frames can theoretically contain either interlaced or progressive material, or even both sequentially, without the equipment which uses it needing to be very aware of what the image data actually represents.
It turns out this can be a bad idea, because interlaced video displays nicely on a CRT but poorly on almost anything else. LCD displays show all of the lines simultaneously, so that the time difference between them can become visible as combing. Often LCD displays will process video they think is interlaced to reduce this. This can be done well but if it's done poorly or unnecessarily it will reduce, perhaps noticeably, the resolution of the video.
Modern equipment often has lots of frame buffers and can drop or duplicate frames to its heart's desire. Many displays negotiate mutually-compatible configurations with the equipment they're connected to, so it's entirely possible that connecting the same camera to two different monitors will yield different signals each time, perhaps including duplicated frames to achieve a compatible signal format.
Well, that became an essay.
P  


More information about the ffmpeg-user mailing list