My (sony) video camera produces MTS-files and can of course be set to different video qualities; the choices with highest picture quality are between 50i or 50p.
The format 50p seems to provide the highest quality possible with my camera (and it produces the largest files). But these video's do not play well on any of our windows- or android devices (jerky image, broken sound) so I conclude this is too much to ask from the CPU. Or is the GPU? Or is it, in fact, the software? I'd like to know.
For the time being I'd be well content to store these video's for later and make do with 25p MPEG-4v2 video instead, as this proved to be a format which plays well on all our devices. But surprisingly '25p' (or 30p for that matter) is not among the choices offered. Surely a device capable of recording 50p movies is also capable of recording 25p movies, so why does it revert to 50i for recording movies in 25 fps?
These theoretical questions aside, my practical question is this:
How do I get the best quality 25p, mp42-encoded, video's from my camera? Would it be by de-interlacing the 50i video? Or is there a program to convert 50p to 25p (for instance by throwing away every odd frame) which gives better results? I am of course aiming for a lossless conversion here.
+ Reply to Thread
Results 1 to 24 of 24
Your playback issues could be either CPU or GPU, depending on which is being tasked to decode the video. A GPU upgrade should solve it, but only in programs that support GPU-accelerated decoding.
A truly lossless conversion isn't possible, unless you want to keep extremely large archive files that can only be played on PC. But I assume you don't mean lossless in the technical sense.
Dropping every other frame is preferable to deinterlacing and a lot faster as well. Note that both methods will produce a video that stutters when compared to a 50p or 50i recording, if the shutter is above 1/25.
In addition, the problem may be neither the CPU or GPU but a player/demuxing problem. MTS files are often difficult to deal with. Try different players with different settings. Try remuxing the video into another container like MKV.
Instead of 1080p25 as your distribution format you might also want to consider 720p50. That is compatible with a lot of devices while still preserving the full temporal resolution. 25 vs. 50 Hz make a much bigger difference than 1080 lines vs. 720 lines. The latter will be unnoticeable to most people.
Thanks for your responses; I learned a lot from them.
Some responses, though, seem somewhat off-topic: I am NOT complaining about video-plaback and I am not about to 'try different players'. On the contrary: my aim is for video which will play well on ALMOST ANY odd system. I found that format in an MP4 container with a H.264 encoded stream of 25p video. Now, I'm just looking for the best way to 'convert' my camera files to this format - hence my question.
Also I'd appreciate it if you could explain some more, especially about "if the shutter is above 1/25"?
The sentence 'both methods will produce a video that stutters when compared to a 50p or 50i recording' I think I do understand up till where you mention 50i. That is where I get lost (not for the first time). Why would 50i display better then de-interlaced 50i, that is 25p? I asked this question elsewhere but never got a clear answer. Most people say a 50i video is always de-interlaced before display. (Never got an explanation about why this is necessary, though.)
But anyway: my question was NOT about stuttering video but about the best way to get to 25p video.
ffmpeg and mp4box) extensively but it didn't seem to make much difference. De-interlacing, however, DID make a big difference. (Sadly, ffmpeg made a mess of de-interlacing my MTS files. I use Freemake now.)
Only with "all intra" codecs (each frame is encoded as a standalone object) could you discard every other frame without reecoding. Like many of the lossless codecs (huffyuv, lagarith, ut), MJPEG, all i-frame MPG, etc.
- Discarding half of the fields, then interpolating
- Interpolating using information from both fields, then discarding half of the result
Most people say a 50i video is always de-interlaced before display. (Never got an explanation about why this is necessary, though.)
But anyway: my question was NOT about stuttering video but about the best way to get to 25p video.
50i is better if it was shot 50i
25p is better if it was shot as 25p
Because both are shot with different shutter speed, for 50i shutter speed is shorter, for 25p longer, but that brings you all kinds of troubles, you do not want to shoot 25p basically , as a home user, difficult to shoot any motion, using handheld camera/camcorder, you'd need ND filter with daylight outdoors. If not, your footage would suck. I shoot 30p and understand it very well what I could shoot. And I have 30p available, cannot imagine having 5 frames gone on the top of it , and using 1/25. You could even use 1/50 without much blur in the video but even that would kind of "strobe" in video while having motions in video or jerky movement. With 1/25 you can barely get things sharp and focused while camera is hand held.
If you shoot 50i, so far so good playing it as is, 50i, TV's handle it (showing 50i to 50p on screen), some devices could have problem deinterlacing it in the best way. But as soon as you start to transfer 50i to 25p you discard temporal resolution, imagine ball coming from left to right, using 50i you'd have 50 times information where it is within a second, 25 p would use only 25 informations about the position where the ball is at the moment, and because shutter was used for 50i, understand shorter, then it would seem jerky on screen. Understand , there is a difference if video is shot as 25p or if video is transcoded from 50i to 25p, you'd get different results. You can also deinterlace 50i to 25p and blend fields, but that ball would be rendered on screen with "ghosts", you could see "quantum physics" effect. The ball would not be there or there, in reality somewhere in the middle, or anywhere you could imagine it in between
You have to transfer 50i to 50p to keep temporal resolution, like TV's do. But why not to shoot it 50p right away then.
You can render, export 1280x720 for tablets, PC's whatever device, you can store both 1920x1080 50p and 1280x720 50p next to each other always and user would decide what to stream.
Last edited by _Al_; 31st Jan 2016 at 14:14.
Ok, I'm getting thoroughly confused now
But let me start with the part that now seems to become a bit more clear to me. vaporean800 wrote "Dropping every other frame is preferable to deinterlacing" and nobody contradicted that. Jagabo contended that this cannot be done losslessly, as codecs work by comparing subsequent frames. Together I take this to mean that a lossy conversion is necessary in any case and that converting from 50p to 25p will give me better results then de-interlacing a 50i video. Am I right??
If so, I only have to find out what conversion program to use for the best results. Any advice on this, anyone?
Just one question keeps nagging me though: when rendering a video, I get to choose between a number of 'zones': e.g. 4 zones (fast) or 1 zone (slower, but better compression). The manual explains this as follows: "using multiple zones, every picture is divided into as many strips, which are compressed independently. This is much faster then having to take the whole picture into account but of course the compression algorithm will be less efficient and the resulting file will be larger." Being in no hurry, I set 'zones' to '1' and indeed, the file got considerably smaller without any loss in quality as far as I could make out. My point here is: there is no mention at all about a setting which would compare the pixels of not just the whole frame but also those from the previous (or the next) one. In fact, I have trouble understanding how comparing with another frame would be even compatible to a setting of '4 zones' considering that the whole speed-gain of multiple 'zones' seems to be that they are independently encoded. I hope Jagabo will care to comment. But anyway, this was just a nagging afterthought.
However, just when I thought I got the whole thing sorted out, other considerations were brought up, notably about 'shutter speed'.
In other words: what is the relation between frame-rate and shutter-speed, if any?
Sure, with a setting of 50p the camera can't use a shutter-speed of 1/25th - that is a limitation to be taken into account when shooting under poor lighting conditions. Apart from that - I can't see what the relation is. And even under poor conditions, I'd expect modern camera's to just produce grainier pictures while maintaining 50 or 60 frames per second. Right or wrong?
Are media-players (I mean computer software) really showing 50i video? Or are they de-interlacing the video on-the-fly and showing 25p instead? And if so, why? And wouldn't it, in that case, be advantageous to start with 25p (or 30p) in the first place?
And what's the difference between a media player and a TV anyway?
Do media-players show a 50p video really as 50p (or 60p as 60p)?
And what about editing: can I edit 50i video without de-interlacing it, that is without loosing 'temporal resolution'?
Again, what you should do:
Record and edit in 1080p50
Use 1080p50 as distribution format if target allows it (like youtube)
Use 720p50 as distribution format if 1080p50 is not allowed on the target player (like BluRay or a slow computer/phone)
Forget about interlacing.
You can deinterlace 50i to 50p, nowadays some become to forget that there was 25p. Anything can play 50p, phones as well. If not 1920x1080 50p, then encode 1280x720 50p.
Forget about that Blu-Ray cannot accept 1920x1080 50p, if this is a case, why would you try to lean on this crippled specs. Blu-Ray specs are not for people but for industry in mind, nobody cares about everyday Joe, they want him buying a new disk , nobody cares about his home video, and why would you authorize it and not just use as data anyway. You can have your movies on thumbdrive, hardisk, a cloud, whatever. Anyway soon you can encode to 50p using UHD Blu-Ray specs, it allows 1920x1080 50p, HEVC perhaps but anyway ...
As soon as you start to compromise like :
deinterlacing 50p or 50i to 25p, what method, for 50p dropping every other frame (jerkiness might be introduced) or blending frames to 25p , 50i to 25i by blending fields, or creating 50p from 50p and then dropping frame or blend frames to 25p , ..
... you compromise a lot
You do not seem to compromise much at all encoding 1280x720 50p from 1920x1080 50p for devices. Try it. Or even 1920x1080 50p but encoded with fast decode or easy settings etc. To deinterlace to 25p is just thing of the past. You can surely think of it that way. That lowered resolution is only version for those tablets anyway. You would not store those 25p deinterlaced versions for good would you? That's too harsh.
TV's , monitor would show 50p, not sure why are you afraid of, if you feed them with 50i, then they have to deinterlace to 50p anyway, they do not show picture like 30 years ago, with fields but with frames. So if there is a half frame coming from interlace footage, they have to construct a whole frame out of it. You got a cheap TV and you get video even with interlace artifacts in it. You have a software platform that plays video in any device that deinterlaces badly, you can see artifacts as well. Stay away from interlace, not like you do not have a choice here. Platforms tend to forget about interlace, I use Kodi to play my old DVavi videos, interlaced, and it cannot properly deinterlace. So not even super TV would help you here if you software puts it on screen. I doubt that they are going to fix that, nobody seems to care about interlace now.
If deinterlacing, the double frame rate deinterlace to 50p, but you are asking 50i or 50p, then 50p. I'd only repeat Sneaker to say it more plainly.
I used my editor (Sony Vegas 13) to 'render' some video's shot at 50p according to my camera settings. First in 50p (a format indicated as 'matching the input') and then an output at 25p. Both in MP4 format.
Result: on my, fairly powerful, desktop the '50p' video played with noticeable jerks where the camera moved fast. On my other devices (laptop, tablet, smartphone) it was even worse.
In contrast, the 25p output played smoothly on all devices.
Just to be sure I repeated the procedure using my other editor (PowerDirector v. 14) and the results were the same or even more pronounced.
Heeding your advice, I subsequently tried 1280x720 at 50p.
The result, to my surprise and possibly yours, was not much different; the 50p video did indeed look somewhat less jerky then 1920x1080 video, but it still didn't look as smooth as the 25p video did, even at 1920x1080. On the other hand, on a 'native 1920x1080' screen, the 1280x720 video looked noticeably less sharp (as I verified with screenshots).
Aside: I'm aware that the quality of displaying 1280x720 on a screen with 1920x1080 pixels may depend on video-hardware, video-drivers and the OS; nevertheless it stands to reason that a genuine 1920x1080 image, which is not alreay cheating somewhere, will look sharper on a 1920x1080 screen.
Summarising: I, for one, have not been able to produce a 50p video which looks better, or even as good as, a 25p (or 30p) video, even when reducing the resolution. How can this be explained - did I make a mistake or are we living on different planets?
It seems to me that 'threads' are a completely different issue, having to do with a division of labour between cpu-cores. This has to do with speed, but nothing with the questions at hand as far as I can see.
The questions I was most curious about thus still seem to remain without answer.
In response to the question of converting 50p to 25p without de-compression/re-compression, Jagabo contended this was impossible because the compression uses pixels form subsequent frames, but didn't answer my objections and describes a process which makes it unlikely that this is the reason.
On the subject of 'interlace' I wondered: do media-players show and interlaced video as such or do they de-interlace the video before showing it? And if so, why? I think I already wrote here or elsewhere on this forum: interlacing is now universally derided, but can anyone explain why? What technological advances make it less desirable now than it was back then?
@sneaker said "For modern screens the playback software (or the display itself) need to de-interlace to either 25p (bad) or 50p (good)" but gave no reasons as to WHY this would be the case. I still don't see any technical necessity for de-interlacing and I don't see any advantage. And again: why wouldn't the reasons to introduce interlacing not be just as valid today?
Then: what is the relation between frame-rate and shutter-speed, if any? The quesion only came up after the posting of _AI_, but I didn't get an answer.
One more: Can I edit 50i video without de-interlacing it, that is without loosing 'temporal resolution'?
The question came up in other threads on this forum and was met with the same hostility toward 'interlacing' but so far I didn't see any valid arguments. What is this aversion against interlacing based on?
I'll throw my 2 cents in...here's why I don't care for interlacing.
Interlacing itself is not always a bad thing. It accomplished it's purpose in years past by reducing bandwidth needs. The problem I find with interlacing is how often it is improperly handled by others. This makes processing the video more challenging. But that's just my opinion.
Sent from my 831C using Tapatalk
It could be wrong settings, project properties, export properties, clips wrongly interpreted by videoeditor (rare), weak PC etc. As always, as always, if you post original clip, 5s, then encoded, 5s, it always clears out. Threads like these can go for millennia, but it always start to make more sense after samples are available.
There are media players nowadays for about $50 that can play UHD in HEVC (not mentioning H.264) in 50p/60p. To think about 25p especially 50i to 25p or 50p to 25p is simply wrong. Tomorrow you'll be sorry.
Last edited by _Al_; 1st Mar 2016 at 18:24.
This thread is ridiculous. OP asks a question, people answer, OP responds with nothing but doubt about the factual accuracy of the answers. Interframe video compression is not some fantasy that jagabo invented.
If 720p50 is still stuttery, you may need to verify whether you have your monitor set to 50Hz instead of 60Hz.
Modern displays are progressive. Even progressive CRT computer monitors completely refresh the screen from top to bottom each refresh cycle, but I'd imagine even if LCDs could be built to refresh the odd scan lines and even scanlines independently, it probably wouldn't look good, and thinking about it, interlaced video would also have to match an interlaced monitor's native resolution. You can't resize interlaced video without de-interlacing it first.
When interlaced video is de-interlaced for a progressive display it's generally de-interlaced by the player/TV to full frame rate.
ie from 25fps interlaced to 50fps progressive to retain fluid of motion. When it's de-interlaced to 25fps progressive it tends to look a little less fluid than natively progressive film at 24fps or 25fps etc, probably due to a lower amount of motion blur (relating to shutter speed).
There's examples in the zip file attached to this post. There's also some more samples attached to post #20. Just a section from a PAL DVD.
One of the challenges of de-interlacing is to take two fields that are different moments in time and combine them into a progressive frame. It's not always easy. Look at the edge of the desk, or the blind on the rear wall at the top right of the frame as the camera pans.
Yadif de-interlacing quality is something like the usual quality of hardware de-interlacing on playback. You should notice a fair difference between Yadif at 25fps and Yadif at 50fps.
QTGMC is a high quality Avisynth de-interlacer. It de-interlaces to 50fps but it can output 25fps when it de-interlaces by dropping every second frame. It can add motion blur to compensate, but 50fps is going to look smoother.
I know nothing about your camera but Level 4.1 support by hardware players is a fairly standard thing for h264 decoding, which means if you want 50fps, you need to stick to 720p. 50fps at 1080p requires Level 4.2 support. https://en.wikipedia.org/wiki/H.264/MPEG-4_AVC#Levels