SD vs HD encoding comparison

[martinfrombern]

i was thinking about this next issue and came up with my explanation to it, however, i would like to hear expert's explanation.
i made an encode of an episode (16:9) from a PAL source without changing resolution at 2500kbps.
i made another encode of the same episode, but this time BluRay sourced, also without changing reolution (full hd) at 12500 kbps.
both encodes were done using the same insane settings (tesa, range 32, ...)
both BluRay and DVD, are superb sources. i suspected, that because full hd contains exactly 5 times as many pixels as PAL, that picture quality of the two encodes should be equally transparent to their respective sources. however, it seems it's not true. second encode (hd) is completely transparent to it's source, but there is some signficant difference between the first one (sd) and its source.
the same thing occurs when i use pal-resized BluRay source instead of dvd pal. i also tested with some other blurays i own and i get similar results (full hd encode is transparent, but pal-resized encode is not transparent with pal-resized source when using 5 times lower bitrate).
i was thinking about an explanation for this and somehow concluded it is due to number of details per pixels, in other words, full hd contains more flat areas which are more compressable.
anybody else thought about this issue? what is expert's opinion? :)

[Selur]

You might want to post your full command line for sd and hd source, some samples, screenshots, etc.,...

[audyovydeo]

i was thinking about this next issue and came up with my explanation to it, however, i would like to hear expert's explanation.
i made an encode of an episode (16:9) from a PAL source without changing resolution at 2500kbps.
i made another encode of the same episode, but this time BluRay sourced, also without changing reolution (full hd) at 12500 kbps.
both encodes were done using the same insane settings (tesa, range 32, ...)
both BluRay and DVD, are superb sources. i suspected, that because full hd contains exactly 5 times as many pixels as PAL, that picture quality of the two encodes should be equally transparent to their respective sources. however, it seems it's not true. second encode (hd) is completely transparent to it's source, but there is some signficant difference between the first one (sd) and its source.
the same thing occurs when i use pal-resized BluRay source instead of dvd pal. i also tested with some other blurays i own and i get similar results (full hd encode is transparent, but pal-resized encode is not transparent with pal-resized source when using 5 times lower bitrate).
i was thinking about an explanation for this and somehow concluded it is due to number of details per pixels, in other words, full hd contains more flat areas which are more compressable.
anybody else thought about this issue? what is expert's opinion? :)


Maybe because you DVD source is almost certainly interlaced, while your BluRay might be progressive ?

cheers
a/v

[nm]

Maybe because you DVD source is almost certainly interlaced, while your BluRay might be progressive ?
PAL DVD movies are rarely interlaced.

[Dark Shikari]

1) The bitrate required for transparency does not scale linearly by number of pixels.

2) When you watch the video, you're (probably) upscaling the video to your full screen size. This means that you see artifacts more clearly with a smaller resolution video, so you need better quality on the lower resolution video to reach transparency.

[Chengbin]

What kind of display are you using to watch these videos?

[G_M_C]

[...] This means that you see artifacts more clearly with a smaller resolution video, so you need better quality on the lower resolution video to reach transparency.

Never thought about this that way. But it is strangely contradictory, but also logical like the logic of a Vulcan :D

[martinfrombern]

1) The bitrate required for transparency does not scale linearly by number of pixels.

2) When you watch the video, you're (probably) upscaling the video to your full screen size. This means that you see artifacts more clearly with a smaller resolution video, so you need better quality on the lower resolution video to reach transparency.

thank you all for your answers. as i said before, i get the same result with when using resized bluray instead of dvd pal (although dvd pal source is also progressive). about dark shikari's comments

1) didn't know that. i mean, i supposed it's not perfectly linear but i was guessing it's something very close to that. i guess i was wrong (from your answer and my tests). is there any approximate connection, or at least is it sublinear or superlinear?

2) i didn't upscale sd when comparing. both encodes were compared at original resolution.

my logic for explaining this was that complexity of sd compared to it's resolution is greater than complexity of hd compared to it's resolution. it seems to me that this should be true, what do you say?

[Dark Shikari]

thank you all for your answers. as i said before, i get the same result with when using resized bluray instead of dvd pal (although dvd pal source is also progressive). about dark shikari's comments

1) didn't know that. i mean, i supposed it's not perfectly linear but i was guessing it's something very close to that. i guess i was wrong (from your answer and my tests). is there any approximate connection, or at least is it sublinear or superlinear?Sublinear with respect to framerate and resolution.my logic for explaining this was that complexity of sd compared to it's resolution is greater than complexity of hd compared to it's resolution. it seems to me that this should be true, what do you say?Yes, this is probably true.

[benwaggoner]

1) didn't know that. i mean, i supposed it's not perfectly linear but i was guessing it's something very close to that. i guess i was wrong (from your answer and my tests). is there any approximate connection, or at least is it sublinear or superlinear?
The rule of thumb has generally been a power of 0.75.

So, if 640x360 looks good at 1000 Kbps, and you wanted to encode at 1280x720, you'd want to start with:

((1280*720)/(640*360))^0.75*1000=2828 Kbps.

YMWV, of course, but if it's not a bad starting point to iterate from.

[Sharktooth]

that depends on the source quality and details.

[Sagekilla]

The scaling looks like (a slight modification on Ben's): bitrate * (R1 / R2) ^ c. Where R1 > R2 (1280x720 and 640x360 for example) and c is a complexity factor. In most cases, it can be approximated to 0.75 as Ben said.

If you're viewing at native res, it's 0.75, but when upscaling is involved the value can be a bit higher. In practice, it's usually not though.

[qyot27]

Is there a similar formula for framerate differences also? Or is it virtually the same as the one for resolution, replacing the resolution variables for framerates instead? I've rendered videos in the past at 24fps or 30fps AND 60fps, which is the main reason I ask.

Also, in the example it was 1280x720 being compared to 640x360 - is the complexity roughly 0.75 when contrasting 1920x1080 to SD sources also, or is it lower still?

[Sagekilla]

There's a similar relationship with framerate, but depends a bit on how good your encoder is. As for other resolutions, the resolution scaling formula does apply still but c is different.

Generally speaking, the smaller the frame the more "details" you have relative to size. Think of a typical scene in a room: Not everything becomes sharper when you capture at a higher resolution. A flat wall may show it's base texture a bit better in high resolution video, but it generally looks identical to an upscaled version of it.