Difference between 720 and 1080 at different bitrates

[tufffta]

Let's assume I encode a 1080p video to 720 at 6000 kbps and to 1080 at 10000 kbps. The difference between the bitrates is less than 2 times (only 1.66). Theoretically 720@6000 would be better than 1080@10000 - would it?
But would it make sense choosing 1080@10000 if I am watching in on 1366x768 screen (meaning 720 video is upscaled and 1080 is downscaled during viewing)?
I understad that depends on a particular video, but how's theoretically?

I know I should just encode samples and see for myself, but the thing is that usually when I see the difference, I don't know which one is better (one may look better to me one time, but the other time the other version may seem better to me).

[frank239]

why not set 720p at 10000 kbps?
The greater the bitrate, video is more clear.

Your screen is suitable for 720p

[pandy]

If frame rate is identical then 1080 can provide slightly higher quality thanks to reduces quantization errors.

[Cornucopia]

Quantization errors?!! No, when the frame rate is identical (meaning 60p or 30p, because there 720i does not exist), for any given bitrate, the clip with the more resolution, while giving more information (usually = more quality) will also have the greater compromise in quantizing, in order to hold the bitrate down.

While not a totally linear (more a logarithmic) process, greater resolution DEMANDS greater bitrate to accommodate it.

However, not all things are usually that equal (framerate, bitrate). 720p @ 6Mbps vs 1080p @ 10Mbps would be hard to compare because one doesn't know where on that logarithmic scale either or both of those points lie. That sort of thing is dependent upon other factors: image complexity (both intra- and inter-frame) is the main one. One would have to do a visual comparison to really know.

Scott

[pandy]

Yes, quantization errors, after downscaling from 1920x1080 to native resolution (this case 1366x768) they will be smaller than for upscaled 1280x720.

[edDV]

Another factor is the quality of the scaler in the TV. If your display is 1366x768, all 1280x720p must be upscaled and all 1080p needs to be downscaled to display resolution. If you want to optimize for your current display, you would encode to display resolution, in this case 1366x768 for pixel by pixel display.

[pandy]

If you want to optimize for your current display, you would encode to display resolution, in this case 1366x768 for pixel by pixel display.

This is not standard HDMI resolution and can be not supported by HDMI Source and HDMI Sink (Source and Sink must support this reoslution and provide necesseary information - EDID - to set this resolution).

Info based on CEA-861-E "Table 4 Video Formats—Video ID Code and Aspect Ratios"

[jagabo]

Forget bitrates and encode with constant quality. You'll always get exactly the quality you specify, regardless of the properties of the video. Encode at the same resolution as your source.

[edDV]

If you want to optimize for your current display, you would encode to display resolution, in this case 1366x768 for pixel by pixel display.

This is not standard HDMI resolution and can be not supported by HDMI Source and HDMI Sink (Source and Sink must support this reoslution and provide necesseary information - EDID - to set this resolution).

Info based on CEA-861-E "Table 4 Video Formats—Video ID Code and Aspect Ratios"

The HDTV manual will list supported resolutions for HDMI and VGA. Many will support 1360-1366x768 for a pixel by pixel playback. Some won't. Most NVidia/ATI cards support 1366x768 out. My main reason for pointing this out was to show picture quality depends on the quality of the TV's scaler. Sending at display resolution avoids hardware scaling.

If your goal is to make an AVCHD or BD disk, then you are limited to 1920x1080, 1440x1080 or 1280x720.

[tufffta]

Forget bitrates and encode with constant quality. You'll always get exactly the quality you specify, regardless of the properties of the video. Encode at the same resolution as your source.

I don't understand it: I get exactly the quality I specify - I mean, how to know WHAT quality I specify by, let's say, using 18? I mean, I don't understand what 18, 20 or any other number means, because these are not percents or anything.

[edDV]

File size depends on input video complexity, noise, etc. You need to test for yourself where your personal quality threshold lies.

18 works out to very good quality while retaining significant compression. Lower causes file size to grow at a greater rate than quality improves.

22 works out to adequate image for small screens, somewhat blocky on big screens.

[Hoser Rob]

Forget bitrates and encode with constant quality. You'll always get exactly the quality you specify, regardless of the properties of the video. Encode at the same resolution as your source.

I don't understand it: I get exactly the quality I specify - I mean, how to know WHAT quality I specify by, let's say, using 18? I mean, I don't understand what 18, 20 or any other number means, because these are not percents or anything.

Unfortunately it's not so simple, despite what all those people who make "fast high quality one click" programs say. eddv is right ... it depends upon source video content/quality and subjective factors.

If you use h.264 encoding, which is the best out there now, constant quality (actually constant rate factor) of 18 to 19 will give you good quality. 19 will yield significantly smaller files but 18 is signigicantly higher quality. I'm using a 23" monitor for watching video ... I don't like much TV, rarely watch it ... and I can clearly see the difference.

Again, it depends on the source. If it's a sundance type dvd with a lot of static scenes I'm not so fussy. If it Lord of the Rings I'm a lot more fussy.

I'm not sure I agree that constant quality really gives you better quality than 2 pass ... none of the program docs I've read claim this, though they claim you can get just as good results. CQ/crf is really target bit rate mode but without an actual target bit rate. What it does do is let you use advanced h.264 settings, which really slow down encoding, and make much better video than you could in the same time as 2 pass without them.

But advanced settings, though they're where the real meat is, are a real kettle of worms for the newbie.

I haven't mentioned one pass target bit rate because I won't touch it with a 10 foot pole. It stinks. The only justification for it is if you're encoding from a streamed/broadcast source. You can't do 2 pass with that. The real reason those "easy fast one click" encoders are so fast and easy is they only have a one pass target br mode. They stink.

[jagabo]

Forget bitrates and encode with constant quality. You'll always get exactly the quality you specify, regardless of the properties of the video. Encode at the same resolution as your source.

I don't understand it: I get exactly the quality I specify - I mean, how to know WHAT quality I specify by, let's say, using 18? I mean, I don't understand what 18, 20 or any other number means, because these are not percents or anything.

Encode a few short, varied, videos with different quantizers. Say, 15, 18, 21, 24. Watch the results and compare to the original. Decide which you would be happy with. You can now use that quantizer for all your encodes because all videos will show the same amount of degradation, regardless of the properties of the individual video.

If you're going to use x264 -- it's weak spot is dark areas with noise:
https://forum.videohelp.com/threads/345427-Handbrake-Should-i-leave-it-on?p=2156674&vie...=1#post2156674

[hello_hello]

I'm not sure I agree that constant quality really gives you better quality than 2 pass ... none of the program docs I've read claim this, though they claim you can get just as good results. CQ/crf is really target bit rate mode but without an actual target bit rate. What it does do is let you use advanced h.264 settings, which really slow down encoding, and make much better video than you could in the same time as 2 pass without them.

According to one of the x264 developers, 2 pass and CRF encoding produce basically the same output, assuming the same bitrate and settings.
http://forum.doom9.org/showthread.php?p=1373437&highlight=ocd#post1373437
"Given the same amount of encoding time, CRF is superior to 2-pass at the same bitrate. Given the same settings rather than the same time, they're effectively identical within margin of error. My recent tests show that CRF generally has a very slight edge, albeit the difference is so small that you'd have to have OCD to care."