640x480 or lesssssss..........

[alx]

This is a very stupid question, but maybe, in some way, interesting.....I´m trying to encode an 1:33 movie (doesn´t matter
the title).....if i made at 640x480, first pass size is about 800 mb., but if i made at 512x384, final size is perfect, near
640 mb., perfect to fit in 1 CD........the question is about what is the best........A) reduce dimensions or B) made a second pass
and reduce quality.
Any suggestions are welcome.
Thanks.
Alx

[ookzDVD]

For 4:3 movies I think 512x... is enough. :)
For 16:9 movies I will never lower than 640x...
Especially if you play it on TV.

[bb]

You'll loose quality either way: either you loose resolution or you get more artefacts / more smoothing (while smoothing also means to loose resolution to some extent).

I prefer the 640xYYY resolutions, because I can play them back at native 640x480 video card resolution (no resizing during playback). Depending on the source you may get better results by reducing the resolution, and the encoding process will be faster (less pixels to encode).

bb

[Steady]

I think it is a judgement call. 640x--- will look better (sharper) in still scenes. 512x--- will look better with lots of motion. Partly it depends on the movie and partly it depends on you. I tend to prefer higher resolution because I like to pick out details in still scenes. I find that if you then to watch movies full screen, low resolution doesn't give much benefit.

[DJ Bobo]

Oh well, the eternal question that I ask myself every day! go for 512x384 if you're using a codec other than DivX 3.11, and 640x480 if you can handle SBC encoding in NanDub

NB: never done an sbc encoding, I feel so ashamed :o
I still feel like NanDub is too complicated for crying out loud!!! :scared:

[^^-+I4004+-^^]

>I think it is a judgement call. 640x--- will look better (sharper) in still scenes. 512x--- will look better with lots of motion

is it really so simple?

lowering the res will lower the sharpness ( and on the motion scenes ) too,but will
yield less blocking because of lower res.(on the same presumed bitrate)


on the other hand....the higher the res , the smaller those
blocks (from ME errors) will be....
(16x16 block is "bigger" on 512x384 than on 720x576..)
(but obviously on relative lo bitrates 720x576 cumulative errors will stack and make image real ugly....)

for best quality i don't resize at all..(720x576) and sure go for 2cd rips....
(and put this to tv-out at 768x576...loking good...)

so:quality-->higher bitrate/higher res.

saving space-->lower bitrate/lower res. (this also means lower quality,off course.....)

fair estimation of 720x576 space eating?
about 50min or so with divx3 on 1 74min cdr........
for some stuff it really pays off.........
and sure some stuff doesn't deserve it..

i woudl say that 576x432 is the lowest you should go if you aim for decent quality...(given that 720x576 is the best it can be...)

cheers

Ivo

[edit:too much blocking/too low bitrate for the res--->ffdshow,addnoise....]

[Justinus]

I didn't prefer to reduce its resolution. First, I resize its height to create a correct ratio. Then, I crop the dark bar around the corner. I use 80min CD with Mode 2 so I can fit almost 800MB on one CD. The quality is very good.

[bb]

Originally posted by Justinus
First, I resize its height to create a correct ratio. Then, I crop the dark bar around the corner.
Would be faster the other way round...

bb

[^^-+I4004+-^^]

why not resize horiz.
it'll probably yield better results then resizing vertical
(even if it means blowing it a bit...)

[Justinus]

I concern its aspect ratio but not up to detail, such as PAR or something similar. I'm still not good at aspect ratio at the moment, so if I say something wrong, just tell me.

@^^-+I4004+-^^
Normally, with 720x576 (I use PAL as an example because this country is using PAL), when it displays, its resolution will be stretched horizontally (e.g. 768x576 for 4:3) in order to restore its original aspect ratio, am I correct? So, I just want to create the correct aspect ratio at the output movie so there is no need to correct it again (I don't intend to watch it through a TV anyway). Thus, instead of resizing horizontally (enlarging), I chose to resizing vertically (reducing). There is no point of enlarging movie when you won't gain anymore quality from the original source but you do when you reduce its size. Moreover, bit distribution seems to be better than enlarging it.

I'm still finding the way to improve this idea. So, any personal idea is welcome :)

@bb
You mean crop first and then resize later. It may be faster than my method. However, if I do that way, then the aspect ratio wouldn't be correct. The area I want to crop may be varied for each movie. You have no idea the net resolution of movie you want to keep. Moreover, after cropping, the net aspect ratio is very difficult to resize into very perfect aspect ratio. In addition, in order to gain the correct aspect ratio, you may need to add black bars or cut of some movie's detail off a few pixels before resizing. On the other hand, if you resize it first, at least you ensure that you've got the right aspect ratio (I'm still not sure that it's correct in technical discipline or not), and then crop it. However, the drawback of this method is it is very difficult to control the final resolution. You won't get what resolution you want, just roughly. Consequently, I choose to have correct aspect ratio rather than desired resolution :D

[^^-+I4004+-^^]

did a small test :

768x576 has AR that closely resembles original tv image
(just compared it on tv-out....i compared 768x576,720x576 and 720x576 with corrected PAR)

720x576 doesn't have correct AR,and if you gonna use wmp only
(and not some player that can correct AR,like zplayer for example)
you'll get video with uncorrect AR ( as wmp doesn't touch AR )
720x576 with corrected PAR (so it's ccir601 compliant) doesn't have
correct AR likewise.....it's just a fraction different from
"normal" bt8x8's 720x576.......


capture at 768x576 and work on this
source to get best quality....

why is vertical more important?
tv system (pal) was made having in mind that 576 lines will always
be there....you can change "precision" of drawing those lines
(ie. lowering the bandwidth ,like VHS does) but you MUST have 576
lines for PAL system to work.....(and VHS has 576 lines)....

so,vertically it is well defined,while horizontally it isn't.....
system is made so that it always has 576 lines while you can tinker with the horizontal resolution (ie. with bandwidth) a bit more....
when you do vertical downsize,you're merging more lines into
one,ie. discarding some of the vertical information....discarding some lines.....this cannot be a good thing.....
when you're doing horizontal resize,you're throwing information too,BUT you're not throwing complete lines, so to speak,but only
some parts of the line.....
( * for sure,the line itself is more important than if it is well drawn line (as in tv-signal) or poorely drawn line (as in VHS,or after resizing it from 768-->480 or 352...ever wondered why SVCD res is 480x576,and CVD is 352x576...hehe).... * )
losing lines hurts more than losing some parts of the line.....
(ever saw a "drop-out" from a VHS tape....doesn't that hurt..hehe)
that's why 352x288 resolution looks so lousy (half of the scanlines lost) and that's why 352x576 still looks pretty well....

a simple test:you have captured to 768x576 and then resized to:
a) 352x576 (202752 pixels)
b) 768x264 (202752 pixels)

both resolutions have same pixel count,but 352x576 looks waay better,
because you have effectively discarded 312 (576-264) lines of "raster" on the "b)" resolution,and raster is there to be rock-solid,to be base on which lines are drawn...
576 lines are a framework,a skeleton of complete video image....
losing one complete line loses much more information than losing 10,20 or 30% of one line.....


------------------

see,this is still a line( and a poor one )




and this is not,it's just an empty space,doesn't represent
anythting.......(this is drop-out on vhs tape or over doing vertical
resizing)




so,keeping it 576 means keeping it sharper,as no lines are lost...only horizontal info is lost,but we can afford that....
the less you resize the better image you have!
resizing to 640x480 and still the image outputted to tv-out will have trouble comparing to mere VHS.....(even with sharpening of ffdshow included,and sophisticated tv-out software...),as VHS effectively has 576 lines....(poorely drawn,but still all the lines are there...)

a simple test is trying to capture PAL on 640x480 resolution (discarding the lines during capture itself.....)
see the results of 720x540 vs. 720x576 lines capturing on
http://i4004.bizhosting.com/
even on 540 the blur is very visible....(surely,VD resizer is not so
rough to simply cut-out the lines like we did in this example,but if you vertically resize enough still the end result is loosing quite a bit of lines.......you cannot put 576 pixels into 480 pixels without loosing some of them.....)

[ what a nice contribution to discussions on the issue of what res. to capture in...hehe ]

i agree with you on issue of resizing first...(that's correct,good way to keep AR..)
and i just love your signature.....(heh)

cheers

ivo

[Justinus]

You make it sound like there's a big different between resizing vertically and horizontally. If so, what is it? :confused:

[^^-+I4004+-^^]

look up.....(edited)

[bb]

The crop first, resize afterwards method is used in GordianKnot for good reason; I recommend to use that program for your size calculations to save you the headache. And yes, you will often crop a few lines of the video (when choosing "Smart Crop"), but you'll get the best compromise regarding aspect ratio correctness. If you capture at 720x576, then resize to some 4:3 resolution, then crop, your aspect ratio will be way off...

For TV playback or interlaced encoding preserving the 576 lines is important, but not for progressive encoding and PC playback. So better forget about the 576 lines stuff in your case.

Brooktree (or Conexant) chipset's ideal capture resolution is 720x576, not 768x576. Don't know where to cite the exact technical specs, but this statement is based on personal experience of hundreds of TV captures. You can find a bunch of posts here agreeing. Never capture at vertical resolutions other than 576 to be sure your video won't be resized through the hardware, else you won't be able to remove phase shift or de-interlace afterwards. If you're lucky the hardware won't resize but just drop some lines when choosing a lower vertical resolution which is close to 576. But that's not what you want either.

Windows media player, e.g. version 6.4, indeed allows to change the aspect ratio. You just have to hold the Shift key while resizing the window. It doesn't allow to enter the aspect ratio by numbers, though.

bb

[Ookami]

@^^-+I4004+-^^

Please, only edit your postings if you're gonna change spelling or so, it's very confusing that you change your theories every now and then... And after that we have n edited postings from you :) . Thanks.

Not to mention that we had these capture resolutions in the capture forum threads AND that your postings are heavily off topic ("the question is about what is the best........A) reduce dimensions or B) made a second pass and reduce quality.").

@bb

Look up Karl's AR for Dummies PDF. You can find it in the capture forum and on VideoXone.

[Justinus]

@^^-+I4004+-^^
Thank for your valuable info. Now, I understand more clearly about resizing.

@Ookami
Sorry for discussing off topic. I'm wrong, too, that I asked ^^-+I4004+-^^ first.

[^^-+I4004+-^^]

>Please, only edit your postings if you're gonna change spelling or so, it's very confusing that you change your theories every now and then... And after that we have n edited postings from you . Thanks.

sometimes i simply must do it....

"theories"?...no they are not anymore...anyone can see for themselves if it's true or not what i have said....just use resizer....



although in this case only pixelAR was a suspicious thing....
and for the most of the industry it still is,so.....


>AND that your postings are heavily off topic ("the question is about what is the best........A) reduce dimensions or B) made a second pass and reduce quality.").

i have already replied to that.....one can choose quality or
compressability.....
now we're just having some fun with AR's.....
(i will conclude my involvement in this thread with this post..)
i hope it's ok to talk AR in this place(?)
as it seems it wasn't discussed enough if GK first crops,then resizes..

>Look up Karl's AR for Dummies PDF. You can find it in the capture forum and on VideoXone

yes,that should help.....
although this thread drifts slowly to "capture","encoding" or "VD,avs" segments,let me just state few more things,suggest few more tools (some not mentioned in "ar for dummies" pdf) that might clear this mess.... (and then thread might be moved to capturing forum...hehe) :

>If you capture at 720x576, then resize to some 4:3 resolution, then crop, your aspect ratio will be way off...

sorry,but completely uncorrect :
if you crop first,then put (resize) that cropped image to 4:3 compatible res,
you're stretching (or squeezing) the image in one or another way --> this means changing the AR !

try it : 720x576->cropbottom 16lines(vhs source),resize to 640x480

compare to

720x576->resize to 640x480,cropbottom 12 lines........

the latter looks better AR wise.....
why?
because removing the pixels (croping) in itself doesn't touch AR,but
if you take that cropped image and resize it after cropping,then you're distoring the image....."boxing" it in something that doesn't fit it....
(as Justinus said it...)
it's an easy test to do..i guarantee you'll notice the AR
offset produced by cropping first.....
(sure....better visible on lower res)

i think GKnot method was made having the speed in mind....
this method distorts AR....
method that i and Justinus propagate keeps the AR....
these are not big distortions(with GK),but are distortions...

>For TV playback or interlaced encoding preserving the 576 lines is important, but not for progressive encoding and PC playback. So better forget about the 576 lines stuff in your case.

i would say that even film source (telecided, ie progressive) will look better on 576 lines then on less res.....
infact i know this for sure as i have made comparissons....
you can't argue that the more you resize the more detail you lose...


>Brooktree (or Conexant) chipset's ideal capture resolution is 720x576, not 768x576

not quite:use VDSync14111,as it has improved bt-tweaker,which can
control the pixelAR....you'll notice that ONLY 768 has the pixelAR
that comes near to "1".....
720 or 704 don't.....they have unsquare pixelAR....
so 720 is not ideal...or perhaps it is if you first correct the pixelAR
with bttweaker of VDSync1411.....(so it gets to "1")

>Don't know where to cite the exact technical specs,

i have read it to some extent....sampling rate for 720 doesn't represent sqaure pixelAR....
768 is much closer....also,video will look a tad better,as 768
is closer to original very high frequency sampled input image...
as i recall original input is well over 1000 (!!!) pixels...
i was right : here's an excerpt from data sheet:

"At a 4*Fsc (CLKx1) sample rate there are 910 pixels for NTSC and 1,135 pixels
for PAL/SECAM within a nominal line time interval (63.5 ms for NTSC and 64 ms
for PAL/SECAM). For square pixel NTSC and PAL/SECAM formats, there
should only be 780 and 944 pixels per video line, respectively. This is because the
square pixel clock rates are slower than a 4*Fsc clock rate; i.e., 12.27 MHz for
NTSC and 14.75 MHz for PAL."

(page 25 of bt878/879 datasheet)

also this:
"The Bt879 requires an 8*Fsc (28.63636 MHz for NTSC and 35.46895 MHz for
PAL) reference time source. The 8*Fsc clock signal, or CLKx2, is divided down to
CLKx1 internally (14.31818 MHz for NTSC and 17.73 MHz for PAL). CLKx2
and CLKx1 are internal signals and are not made available to the system.
UltraLock™ operates at CLKx1 although the input waveform is sampled at
CLKx2 then low pass filtered and decimated to CLKx1 sample rate."

yeap,this beast outperforms ADVC100 if huffyuv is used as input compressor....it is originally sampling at 35,4MHZ the "input wave form" (video)
for PAL.....impressive stuff !

(bthardvare resizer is for sure worse than precise bicubic of VD or avs...so,get as many pixels as you can....or bt resizer will do it's resizing....)


> but this statement is based on personal experience of hundreds of TV captures.

watched with unaltered AR of wmp6.4 (i never do that trick you mentioned.....dragging the window to resize etc.) ie. with original AR
720x576 doesn't fill the 4:3 of monitor, ergo it has uncorrect AR...

a bit lenghty,but i hope it helps to clear out some
things.....
(doing tests for yourselves is even better....)


/ivo
[over&out]

[Ookami]

@^^-+I4004+-^^

First, I don't have anything against off topic discussions, in fact discussions are always alive and tend to jump, more or less, from one point to another.

But, somewhere you have to draw the line. For instance, instead of editing/replying in very lenghty postings (nothing wrong with that, BTW; thanks for your motivation), you could've:

a) posted a few links from the capture forum (FAQ, the "PAL resolution is..." thread etc.)
b) made one explanation that stays + links
c) opened a new thread in the capture forum (as you seem to have the motivation to discuss already discussed topics)
and so on.

I have noted several mistakes in your first posting (altough I don't know if you've corrected that, as you tend to edit and re-edit your postings...), therefore I've called them "theories". If you would not constantly change your initial theories/postings, people could much easier discuss/write counter-arguments.

But, if the initial argument changes constantly, where's the need to counter-argument? I for one, don't have the time and motivation to "fight" against theories/postings wich change constantly... Altough, I don't have the motivation to "fight" even against 100% theories that are not edited, so...

Er, who cares anyway what I think? :D

Not to mention that you and bb are somewhat talking about different things! For instance:

bb: "For TV playback or interlaced encoding preserving the 576 lines is important, but not for progressive encoding and PC playback. So better forget about the 576 lines stuff in your case."

And so on.

To make a end to this very lenghty posting, please, try to be on topic in the future threads it would really help.

As the original question(s) is more or less answered, you can discuss this further in the direction it's heading (if you ask me, anyway)...

Cheers,

Mijo.