Why would I want to scale?

[peter42]

I red a lot of articles, instructions and so on about high quality rips and so on but one question is still left. Why would I want to scale?

My guess is that it makes only sense to scale the movie if you want to play it on a standard display like old TVs with an aspect ratio of 4:3 or on newer ones like 16:9.

I want to display the rip only on my own displays which have all other aspect ratios. For example my notebook has a resolution of 1440x900, my external monitor has a resolution of 1280x1024.

In the end my question is, does it makes sense to scale a movie for modern displays? I mean e.g. in 5 to 10 years I probably have a dozens of different screens with different aspect ratios (TV, notebook, external monitor, iPod, ...). Therefore, I guess it does not make sense to scale a movie if I want to play it on different screens. What are your thoughts about this?

And a second and last question. Scaling means just changing the aspect ratio if you do not scale both axes with the same amount. So if I scale only one dimension the result might look a bit squeezed if I scale to much, right?

Cheers,
Peter

[Ghitulescu]

You don't want to :p

1. A high-quality rip is the 1:1 copy of the original. There are however exceptions, when the starting material is of a subpar quality that begs for improvements.

2. Scaling is needed only when the player/display is not able to do this or does it badly or simply defective. There is also an exception to this, when the input video is 43LB and the viewer wants it as 16:9 (crop and scale).

[peter42]

1. A high-quality rip is the 1:1 copy of the original. There are however exceptions, when the starting material is of a subpar quality that begs for improvements.

What exceptions could this be? I can barely imagine how I could come up with a better quality of a movie by scaling.

Thanks for your answers so far! Now I feel more comfortable ripping my DVD collection and not wasting my CPU cycles ;-)

Regards,
Peter

[poisondeathray]

What exceptions could this be? I can barely imagine how I could come up with a better quality of a movie by scaling.

I think he was referring to restoration efforts. e.g. The original source isn't very good. It might be too noisy, or colors are off, etc.... Then the percieved quality can often be improved dramatically by selective filtering instead of doing a 1:1 identical rip.

[Blue_MiSfit]

Correct.

Things like hard telecine, dot crawl, rainbows, strange colors on black and white footage, and widescreen video letterboxed to 4x3 etc can all be cleaned up and produce a result that's subjectively "better" than the source.

Derek

[peter42]

Yeah that makes sense to me.

Just to be 100% sure ;-) Here is another setup where I would say that scaling makes sense just to full fill all recommendations:

Source: Interlaced PAL 720x576
Destination: H.264 with AAC and a MP4 container

The movie has some black borders at the left and right side. Mplayer with crop detection (-vf cropdetect) says that I should crop 10 pixels but it does not know that it is an interlaced movie. 10 ist not divisable by 4. Therefore, I should crop 12 pixels but then the result wouldn't be divisable by 16 anymore: (720-12-12) / 16 = 43.5

I could solve this problem by cropping 16 pixels but I don't like this because then I would crop too much. To solve this issue I could scale the movie. The question is up or down?

a) scale up to 704
or
b) scale down to 688

Did I miss something during all these steps? What would you prefer, scale up or down? Scaling about 8 pixels doesn't make the movie look strange/squeezed or does it?

Regards,
Peter

[smok3]

i would crop 32 pixels in width (no scaling), it is better to crop a bit more than to scale (in my humble opinion):

# Crop(Left, top, -right, -Bottom)
Crop(12,0,-20,-0) # 688x576 (modw16,modh16)
# x264 --sar 1067:1000

then encode with x264 with appropriate --sar flag (i'am guessing this is 4:3 stuff?)

http://someotherstuff.co.cc/resize/index.php?ssmw=720&sar=1.06667&sar2=&ssmh=576&CT=&CL=12&CR=20&CB=&mplayCrop=&trw=&dar=1.06667&dar2=&modw=&modh=&padw=&padh=&css=&doit=true

[peter42]

i'am guessing this is 4:3 stuff?

Yeah right.

OK, I did a sample rip of 10 minutes and encounter that it is a poor quality movie. Not because of my rip, instead what I encountered is that the very first pixel row is half black and half coloured at top and bottom. So I will have to crop that on too (who created such a crappy disc in the first place? ;-)).

As you suggested it might be better to crop a bit more than scaling I use the following setup:

crop left and right = 16
crop top and bottom = 8

So I end up with a movie of 688x560 pixels. Cropping parameters are divisible by 4 and the movie width and height are divisible by 16.

Thanks for all your answers! Hopefully I will have a high quality rip this afternoon :)

Regards,
Peter

[AnonCrow]

the very first pixel row is half black and half coloured at top and bottom.
Perfectly normal and expected on certain types of sources.

[peter42]

Perfectly normal and expected on certain types of sources.

Really? But why? I mean I haven't realized it before but now since I know it, it looks strange to me ;-) Especially if I switch to fullscreen mode the "row" gets bigger and looks kind of messed up. Is there some rationality behind it or was the DVD creator sleeping ;-) ?

[Guest]

The analog spec is the source of it (http://www.futuretech.blinkenlights.nl/resvid.html):

"For NTSC, the first (odd) field starts with a whole line and ends with a half line; the second (even) field starts with a half line and ends with a whole line. Each NTSC field contains 242.5 active lines and 20 lines of vertical blanking.

Similarly, for PAL, the first (even) field starts with a half line and ends with a whole line; the second (odd) field starts with a whole line and ends with a half line. Each PAL field contains 287.5 active lines and 25 lines of vertical blanking."

So if the DVD is created naively from an analog source, then you see the half lines.

[peter42]

Oh OK. Haven't thought that there are so many differences, ifs and thens etc. Doing a high quality rip is no magic but close to it ;-)

Nevertheless, thanks to all of you for your answers! I'm still looking forward to my first rip.

Regards,
Peter

[Ghitulescu]

What exceptions could this be? I can barely imagine how I could come up with a better quality of a movie by scaling.

I think he was referring to restoration efforts. e.g. The original source isn't very good. It might be too noisy, or colors are off, etc.... Then the percieved quality can often be improved dramatically by selective filtering instead of doing a 1:1 identical rip.

Things like hard telecine, dot crawl, rainbows, strange colors on black and white footage, and widescreen video letterboxed to 4x3 etc can all be cleaned up and produce a result that's subjectively "better" than the source.
Yes, this is what I meant.

Concerning the scaling, most people do not have expensive scalers (a good scaler, in my view, cost in 2009 some 1000+€ or at least 500€ if embedded, 2005 they cost 3x more) so they have always quality issues by upscaling SD material. They may want to use software algorithms to overcome the hardware deficiencies.
So, in the end, it's a personal decision to choose between spending a certain amount of money on a hardware upscaler or another amount of money on the electricity bill (software upscaling + recoding).

Not because of my rip, instead what I encountered is that the very first pixel row is half black and half coloured at top and bottom. So I will have to crop that on too (who created such a crappy disc in the first place? ;-)).
Walter Bruch is guilty as charged, as he picked up a National Television System Committee substandard and created the PAL standard in 1963. Halflines are part of both standards, PAL and NTSC.

[Mug Funky]

those half-lines... with 4:3 stuff there's no need to crop it off for TV land, so it seldom is. in 16:9 pictures displayed in 4:3 the lines show up because they are pushed into the safe areas (anything outside 10% from the edge of the screen is usually ignored in TV land, often even boom mics and other on-set mistakes).

some TV stations (SBS australia) will even reject a master that has these lines blanked, citing that there's picture missing.

in DVD land, usually they're blanked so there's a nice edge there - too many complaints from clients and customers otherwise.

[peter42]

# x264 --sar 1067:1000

then encode with x264 with appropriate --sar flag (i'am guessing this is 4:3 stuff?)


Uff I managed to over read that I should have specified the --sar flag. My ripping log files say that x264 automagically decided to use a sar value of 733/688 which is close to yours:

1067/1000 = 1.067
733/688 ~ 1.065

Does this mean that I can throw away all my rips I have done over the last week?

My understanding of the sar value is, that it is a flag in the container format which specifies how the decoder should transform the blocks. If I interpret this correctly then I can still use my encodings, I just have to adjust the sar value inside the container format. That means, the sar value is not used during encoding, but it is used by the player/decoder. I just wonder why x264 needs it then?

Did I get it right?

Regards,
Peter

[poisondeathray]

My understanding of the sar value is, that it is a flag in the container format which specifies how the decoder should transform the blocks. If I interpret this correctly then I can still use my encodings, I just have to adjust the sar value inside the container format. That means, the sar value is not used during encoding, but it is used by the player/decoder. I just wonder why x264 needs it then?



--sar specifies values that will be encoded into the bitstream.

This is different than container level signalling. e.g. you could specify a DAR in mkvmerge or mp4box. The container level signalling usually overrides any SAR values in the bitstream