Which would be the better quality????....Cant decide

[DrRaven]

Hi Everyone,

Apologies if I have posted this to the wrong section....just couldnt work out where to post.

I've inherited a number of video files from a number of people and many of the files are of the same content...but with varying parameters. So...given the following have the same content...

File 1 : 512 x 270, 23.976 fps, 854.9 kbps, Qf: 0.258 bits/pixel (1:32:43)
File 2 : 640 x 336, 23.976 fps, 891.6 kbps, Qf: 0.173 bits/pixel (1:32:42)
File 3 : 616 x 320, 23.976 fps, 894.2 kbps, Qf: 0.189 bits/pixel (1:32:01)
File 4 : 608 x 336, 23.976 fps, 1006.9 kbps, Qf: 0.206 bits/pixel (1:31:19)

Which one will likely have the best quality?....and which one will likely display the best on a large screen TV?

For the purpose of the discussion assume that the same video codec has been used.

thanks everyone....

Brant

[jagabo]

The one that looks best. The rest is meaningless.

[DrRaven]

Hi jagabo,

While I would agree with your response...I simply dont have the time to sit and compare the same file for viewable qaulity...besides...I dont actually have a large screen TV.

The must surely be some indicative metrics that will at least point in the correct direction??

No??

Brant

[jagabo]

If they were all made by the same person, from the same source, with the same program, with the same codec, and similar settings, things like Qf might mean something. But since that's not the case Qf means nothing more that the average bits per pixel. For example, 512x270 with a low Qf from a DVD will look much better than 640x336 with a high Qf from VHS tape.

[TimA-C]

How long does it take to watch a few brief snippets from 4 clips? Probably a lot less time than you've spent faffing around here.

[DrRaven]

Dude....

Its not faffing as you put it....its about discovering if there is a general way of determining quality....and its more than four clips..

Brant

[jagabo]

The only definitive way to determine video quality is to look at it. Period. Look at the "clean" and "noisy" videos in the following post:

https://forum.videohelp.com/topic357646-60.html#1903520

GSpot will tell you the Qf of the clean video is 0.049, the Qf of the noisy video is 1.021. Which has the better visual quality?

[PuzZLeR]

Ok, here's my "faffin'"... :P

Much depends on the complexity of the source itself and your subjective interpretation of the results. You can't fit the same clothing size on all human beings and, even so, you can't have the same taste for all styles either (my thinking relative to Jagabo's).

But, assuming the same codec and the same source, and ONLY assuming as such, a good indicator would indeed be BPP (bits per pixel), but even so, it's not so straight-forward.

This is only assuming the BPP you use is not overkill to begin with. If you were to plot quality as a function of bitrate on a graph you would reach a certain point where it would asymptotically be near a certain quality level no matter how high you raise the bitrate.

If you have a rez of 320x240 and a BPP=4, then it likely would not look any better than a clip of the same rez at BPP=3 in most cases. Thus, you may then want to enlarge the resolution.

So, having said that, if want a rough idea and an opinion on optimizing your output per BPP, here's how I look at it as a theoretical model.

Start with a small resolution and raise the bitrate incrementally. When you reach the point of high quality (near overkill), start raising the resolution and raise the bitrate incrementally again. Keep doing this until the law of diminishing returns applies to the source itself. Once you reach this point, from what your source provides you with, you have achieved the best, and most efficient, possible output for the big screen - lowest BPP for maximum resolution - which would be my objective in your case.

Unfortunately, it's not so simple. If there's a tool out there that can determine the complexity of a video beforehand without re-encoding, please someone post it. I would be more than happy to assist with providing a formula for such a question then.

[edDV]

File 1 : 512 x 270, 23.976 fps, 854.9 kbps, Qf: 0.258 bits/pixel (1:32:43)
File 2 : 640 x 336, 23.976 fps, 891.6 kbps, Qf: 0.173 bits/pixel (1:32:42)
File 3 : 616 x 320, 23.976 fps, 894.2 kbps, Qf: 0.189 bits/pixel (1:32:01)
File 4 : 608 x 336, 23.976 fps, 1006.9 kbps, Qf: 0.206 bits/pixel (1:31:19)

Also these are all non-standard frame sizes so would give unpredictable resizing and upscale quality if fed to a digital SD/HD TV. Standard SD "PAL" frame sizes are

352x288 4:3 (aka CIF)
352x576 4:3 (aka half D1)*
704x576 4:3 or 16:9 *
720x576 4:3 or 16:9 *

* non-square pixels

[DarrellS]

And none of them have a high enough bitrate to look good on a medium to large screen TV. That's the problem with most files that are downloaded with Limewire or Bearshare.

Just one more thing wrong with the RIAA. Suing people for downloading and sharing crap that no-one in their right mind would pay for in the first place.

[tomlee59]

Dude....

Its not faffing as you put it....its about discovering if there is a general way of determining quality....and its more than four clips..

Brant

The problem is that human perception is subjective, but you are seeking a simple, single-dimensional objective metric. These two simply do not map into each other.

Objective metrics do exist (such as PSNR), but they are only rough guides. If simple, reliable objective metrics did exist, ccodec design would be trivial. Unfortunately, huge amounts of testing with human observers remains a necessity. Consider how much was involved in developig MP3 -- this audio codec took years of development. Testing with humans revealed the lowest perceived quality was with German male voices (ironic because much of the development was done at Fraunhofer in Germany), despite the relatively low PSNR vaues.

Knowing these facts is why the others here have said that looking at clips is really the only reliable way to accomplish what you originally asked. And a different person would likely rate quality differently from you. Quality -- as beauty -- really is in the eye of the beholder.

[jagabo]

And or course, PSNR requires that you have the original source to compare to.

[DrRaven]

Thank you everyone for posting.

When I asked my question I had suspected there was no simple method...I guess I was hoping for a simple formulaic answer.

Oh well....we live in hope.

Thank you again everyone.

Brant

[PuzZLeR]

I guess I was hoping for a simple formulaic answer.

Well, like I said, if someone knows a tool that can determine the complexity of a video beforehand (using some "index" given a relative frame of reference) I can surely derive a formula (with some help from the forum of course).