resolution too low?

[Tribun]

hi ppl! i'm trying to encode a movie for which i got a resolution of w512 x something in gordian knot as i tried to have the bits/(pixelxframe) value of about 0,17. but the compressibility check returned such a value that i'd have to decrease the res to about w320 x something in order to be in 40-50%. isn't that too low res? perhaps the only way i should go is increase file size right? thanks for comments

[niamh]

isn't that too low res? perhaps the only way i should go is increase file size right?
Yes and yes :)

[cyberVera]

Please don't get it as an inevitable thing :-)
Some people easily ignore what the compressibilty check says - they just rip and compress it to 700Mb.
You might want to sacrifice audio quality as well, or to cut off credits.

[ukb008]

Hi.

Yes, Width 320 is low, but what's seriously wrong with it? What's the cut-off figure from normal to Low?

Viewing on a TV it might not matter too much. On projection to an 100" diagonal it might look a little blurry, though.

Regards.

[cyberVera]

Originally posted by ukb008
Hi.

Yes, Width 320 is low, but what's seriously wrong with it? What's the cut-off figure from normal to Low?



Resizing to pc's fullscreen will make it very blurry.
For me, minimal acceptable is something around 480-512.

I wonder what was the length of the movie Tribun was speaking about and what were audio settings...

[Tribun]

Originally posted by cyberVera
I wonder what was the length of the movie Tribun was speaking about and what were audio settings...

well it's a movie ripped from tv in dvd format (720x576, avg. 6mbps bitrate) 55min long and audio mp3 128kbps. it's not very long but perhaps hardly compressible :-(

anyway the question about res was more general. now you confirmed the solutions i wasn't sure about :-)

one more thing that i wonder about is how do i estimate in the beginning my target resolution for the movie? is it the few examples doom9 recommends for different aspect ratios of movies? however the the result of compress check that i do later usually requires to change the resolution regardless of my first guess.

[niamh]

Well 2 things.
Either you want a fixed resolution, and you have to adapt your filesize to it
Or you want a fixed filesize and you'll need to adapt the resolution to it (within reasonable bounds, that is)

That's assuming you have good eyesight and a notion of what good quality is(entirely relative); It's easy to produce ugly trash at whatever size/ resolution.

The things you need to consider are
-if the resolution is too low, then when you play back, it will look ugly because there aren't enough pixels altogether.
-if the size is too small, then it will look ugly because there aren't enough bits of data available to fill the screen nicely.
If you're going to watch on an ordinary television,you can get away with a lower resolution as a rule as well. For PC playback, forget it.
You can let GKnot decide the resolutions available to you, it's the simplest safe way.

TV captures are notoriously hard to encode due to all the stray noise and low quality, I suggest you start looking into avisynth filters to increase compressibility as well.

[cyberVera]

@niamh
Sorry, Lady. Not 2 things, but 3 :-)

All that idea about compressibility check is absolutely misleading. There are no logical arguements in using it.
It is pretty simple.

You have three value to deal with:
1. File size
2. Frame size (resolution)
3. Image quality
And you HAVE to sacrifice one of this values.

First, you have to decide what target file size you want - 1 or 2 CDs, or a part of DVD (700Mb, 1400Mb, etc.). I do not see any point in encoding to 2CDs - what for? to wait when the time to change cds will come? or to get "better" quality? If you need maximum "better" quality just buy a DVD - the same money and no efforts. As for me, I just prefer to watch a movie, not to analyse which microblock in which frame is more microblocky than others.
Then, you decide what is more important to you - either a big frame, or a better quality. Here it is not as simple as it seems. Small resolution (for example 320 px width) will make quality of a frame higher. But you will loose all that as soon as you stretch your picture fullscreen. Minimum acceptable here is 480 px, maximum - 640, optimal - 512. Or, if you choose high resolution (640-720), you'll have lower image quality AND bigger load on your pc/videocard.
The third value (image quality) is foredoomed9 :-) You can not do anything with it. Sad but true. All wisdoms of this forum will only teach you to sacrifice a bit of each of the values, but you will end up with the same.

And please, Tribun, do not let that encoding "disease" infect you - total sum of your efforts will be the same, whatever value you sacrifice. Just choose a media and resolution. Forget about compressibility check. And of course choose a codec you like more than others!

Thats all you can do, as a human :-)

[ukb008]

Peace be on mankind, Amen.

Regards.

[jggimi]

I disagree about compressibility, CyberVera...unless I misunderstood you?

I find that content compressibility does impact my decisions about both resolution and framesize. I can compare two different films that have the same aspect ratio and the same length... yet produce widely diverse subjective quality if I use the same resolutions and bitrate.

SPR, for example, is a fairly uncompressible film -- compare a length of it with the same length of any other 1.85:1 film at the same resolution and bitrate (same b/p*f), and I am sure you'll see a dramatic difference in subjective image quality.

In the GK guides, Doom9 recommends some b/p*f values -- as starting points, only, prior to the first compressibility test. After that, they should be ignored.

The value one uses to make one's decisions (or, if using AGK, the value it uses) is a percentage bitrate of the maximum b/p*f possible with that particular content at that selected resolution. One doesn't use a specific b/p*f.

You don't have to use compressibility testing, CyberVera, but I will continue to do so, and I'll continue to recommend it to GK users.

[cyberVera]

I'm afraid I failed to explain it correctly.

1. Resolution and framesize, in given context, are the same things. Framesize = width x height.

2. Two DIFFERENT films! Almost deffinetly two different movies have different compressibilty (they are different, not identical :-)

But what can you do with one given movie? You WILL sacrifice one value for the sake of two others. And you do not need compressibility check to choose which one to sacrifice.

Finally, whatever way you choose, you will end up with the same maximum possible fullscreen image quality.

[jggimi]

Got it, thanks for the explanation!

[ukb008]

For a given filesize, picture quality varies only with framesize. Okay?
That presupposes there are no other factors in this scheme of things, are there?

I feel there's a fourth factor, but for a given encode that remains the same. I mean the codec - the other important factor that can seriously matter.

So to sum up:

Quality (the objective aspects) depends on:

1. File-size
2. Codec
3. Frame-size
4. Source material (that will give rise to the compressibility factor)

Anything else, PROs?

Regards

[cyberVera]

Right.

But codec was mentioned ...
Source influence is obvious and you can not change it. Consider it a constanta. You can not sacrifice it. Otherwise you will continue including audio bitrate, videocard driver, monitor profile, etc.
Quality is a factor of itself, and it is always objective.

Frankly speaking I am not sure about the following.
So, we have 1 CD. DivX codec (or XviD, if you like).
And now, what gives better overall FULLSCREEN impression
480px width output or 640px output for the SAME movie?
One one hand 640 is less blurry in fullscreen than 480,
on the other hand it received less bits per pixel.
Is it a compromise between bluriness and microblocks or something else?

I would appreciate your experience/opinion.

[ukb008]

By 'Fullscreen' do you mean 4:3 images, filling up the traditional TV screen? (Because that's the usual idea of fullscreen).

And why should a width of 640 receive less bits per pixel than a width of 480 as a rule of thumb?

Regards.

[cyberVera]

do you mean 4:3 images

No, not necessarily, just an avi switched to fullscreen.

why should a width of 640 receive less bits per pixel than a width of 480

Because Bits/(Pixel*Frame) depends on total "square" of a frame (= width x height).

[Tribun]

Originally posted by cyberVera
Then, you decide what is more important to you - either a big frame, or a better quality. Here it is not as simple as it seems. Small resolution (for example 320 px width) will make quality of a frame higher. But you will loose all that as soon as you stretch your picture fullscreen. Minimum acceptable here is 480 px, maximum - 640, optimal - 512. Or, if you choose high resolution (640-720), you'll have lower image quality AND bigger load on your pc/videocard.
Originally posted by jggimi
The value one uses to make one's decisions (or, if using AGK, the value it uses) is a percentage bitrate of the maximum b/p*f possible with that particular content at that selected resolution. One doesn't use a specific b/p*f.


i'm desperately trying to understand this ;-) regarding the above post and some other can i simply encode every movie (4:3 as well as 16:9) with 480-640 px width whichever would have the smallest aspect ratio error (considering within this interval quality is guaranteed) and then work with the compressibility check result to select an optimal file size for the movie? that is at this final stage i'll either decrease or increase the target filesize to get around 50% of the test value and i'm done?

[cyberVera]

I give up.

[ukb008]

Why?

[BigDid]

Originally posted by cyberVera
I give up.
:rolleyes:
Sleep on it,
Take your time,
Re-examine (not read but scrutinize) the whole thread,
Question yourself: what is(are) the question(s), what answers were given, can it be sorted/re-formulated?

I've been into it before and out with patience and effort(s) :)

Did

@ ukb008/OT
I am also south of the equator, south of Hawaii ... and you?

[Tribun]

Originally posted by cyberVera
I give up.

was my last post really so frightening :scared: and i thought i've finally found my future universal ripping strategy :D i decided to use auto gknot for this one and see what settings it uses and compare the results with my choice in the "manual" gknot ;)

[cyberVera]

Guys, I'm sorry for not replying for so long. The reason is banal - I was dead drunk due to the victory in the World War II (In Russia it is one of the most respected holidays).

@ Tribun

was my last post really so frightening

Yes, for me it was. Despite my efforts to explain myself I realised that "compressibility-wise" idea was the only one you were thinking about.
Your idea to try to encode a sample video with AutoGK and then with GK is really worth trying. I tried such thing myself and it was a valuable experience.

i thought i've finally found my future universal ripping strategy

I'm afraid it is just a dream. As soon as you realise that you have such strategy you realise that it can be done better. That is why we are members of this forum. Your own experience and trials together with global experience and trials is the right way.

Back to strategy, as I see it (the way I do it):
1 CD
(quote) I do not see any point in encoding to 2CDs - what for? to wait when the time to change cds will come? or to get "better" quality? If you need maximum "better" quality just buy a DVD - the same money and no efforts. So, 700Mb.
DivX
I prefer DivX instead of XviD. DivX is less hungry for pc resources and less buggy. Also it sticks closer to target filesize.
Audio
128kbps ABR for movies less than 2 hours, 112 for more than 2 hours, 96 for more than 2 hours 30 minutes.
Resolution
I always use 512px width. Height depends on source's height of course.
I never perform compressibilty check Do not see any reason why I should. What will it change? Let's say this check says 40. So what? Will I decrease width to 480? No. Because after resizing to fullscreen view I'll get more soapy (blurry) picture. Why not to use 640 if compressibility permits? Because I'll get more pixels in every frame, thus more load on pc resources. It's ok with my machines, but I often give CDs to my friends and they are still satisfied with Pentium 3 or even 2. Also bigger framesize means less quality, as every frame receives less amount pixelwise.
AutoGK (with fixed frame width mode) runs compressibility check just to decide which resizing method to use lanczos or bicubic (if it is low). I always prefer lanczos over bicubic, so it is useless to me.

It is just my point of view. Also, if you really want to understand it better, read again my posts. I can not explain it better. Make your own decision and it will be your optimal strategy.

[Tribun]

now this is quite clear and simple :)