x264 - profiles and logs

[opti2k4]

My graduate work at my university is about h264. Now for practical part of work i have 1080p mpeg2 .ts with bitrate 17-19Mbit. My professor told me to make x264 coding with baseline,main and HQ profile and bitrates 9,4.5,2,1,0.5,0.25Mbit in coder settings. Now i have logs for every coding and turns out those H264 profiles (baseline, main and HQ) are almost the same according to log. I used predefined coding configs in MEGui (CE-Baseline, CE-Main, HQ-slowest). Here are 2nd pass logs for 1000kbit bitrate for all 3 profiles.

BASE
avis [info]: 1280x720 @ 23.98 fps (1236 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSSE3
x264 [warning]: VBV maxrate specified, but no bufsize.
x264 [info]: slice I:15 Avg QP:33.40 size: 30925 PSNR Mean Y:35.92
U:42.62 V:43.66 Avg:37.24 Global:36.55
x264 [info]: slice P:1221 Avg QP:36.23 size: 4916 PSNR Mean Y:33.56
U:41.69 V:42.96 Avg:35.01 Global:34.32
x264 [info]: mb I I16..4: 60.3% 0.0% 39.7%
x264 [info]: mb P I16..4: 9.7% 0.0% 1.8% P16..4: 23.5% 4.2% 0.2%
0.0% 0.0% skip:60.6%
x264 [info]: SSIM Mean Y:0.8688881
x264 [info]: PSNR Mean Y:33.586 U:41.697 V:42.971 Avg:35.039 Global:34.338
kb/s:1003.41

encoded 1236 frames, 9.26 fps, 1003.57 kb/s
desired video bitrate of this job: 1000 kbit/s - obtained video bitrate
(approximate): 1006 kbit/s



HIGH
avis [info]: 1280x720 @ 23.98 fps (1236 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSSE3
x264 [warning]: VBV maxrate specified, but no bufsize.
x264 [info]: slice I:15 Avg QP:33.07 size: 30922 PSNR Mean Y:36.34 U:42.78 V:43.85 Avg:37.65 Global:37.05
x264 [info]: slice P:647 Avg QP:36.23 size: 7656 PSNR Mean Y:35.05 U:42.12 V:43.34 Avg:36.42 Global:35.68
x264 [info]: slice B:574 Avg QP:37.54 size: 1860 PSNR Mean Y:33.42 U:41.78 V:43.35 Avg:34.90 Global:34.36
x264 [info]: mb I I16..4: 46.1% 41.4% 12.5%
x264 [info]: mb P I16..4: 16.0% 4.1% 0.7% P16..4: 36.6% 3.7% 0.5% 0.0% 0.0% skip:38.4%
x264 [info]: mb B I16..4: 0.5% 0.3% 0.1% B16..8: 15.5% 0.6% 0.5% direct: 0.4% skip:82.1%
x264 [info]: 8x8 transform intra:22.1% inter:51.3%
x264 [info]: direct mvs spatial:61.0% temporal:39.0%
x264 [info]: ref P 66.6% 14.9% 8.4% 2.1% 2.1% 1.9% 1.9% 0.7% 0.7% 0.8%
x264 [info]: ref B 82.4% 8.4% 3.4% 1.4% 1.2% 1.0% 0.9% 0.5% 0.4% 0.5%
x264 [info]: SSIM Mean Y:0.8833991
x264 [info]: PSNR Mean Y:34.306 U:41.966 V:43.350 Avg:35.729 Global:35.028 kb/s:1006.40

encoded 1236 frames, 3.58 fps, 1006.58 kb/s
desired video bitrate of this job: 1000 kbit/s - obtained video bitrate (approximate): 1009 kbit/s



MAIN
avis [info]: 1280x720 @ 23.98 fps (1236 frames)
x264 [info]: using cpu capabilities: MMX MMXEXT SSE SSE2 SSSE3
x264 [warning]: VBV maxrate specified, but no bufsize.
x264 [info]: slice I:15 Avg QP:31.73 size: 33798 PSNR Mean Y:36.73
U:43.10 V:44.26 Avg:38.03 Global:37.44
x264 [info]: slice P:647 Avg QP:34.90 size: 7219 PSNR Mean Y:35.07
U:42.37 V:43.60 Avg:36.46 Global:35.75
x264 [info]: slice B:574 Avg QP:36.16 size: 2278 PSNR Mean Y:33.72
U:42.02 V:43.61 Avg:35.20 Global:34.65
x264 [info]: mb I I16..4: 64.9% 0.0% 35.1%
x264 [info]: mb P I16..4: 18.7% 0.0% 3.4% P16..4: 24.4% 3.3% 0.7%
0.0% 0.0% skip:49.4%
x264 [info]: mb B I16..4: 0.5% 0.0% 0.2% B16..8: 22.5% 0.6% 1.1%
direct: 0.4% skip:74.7%
x264 [info]: ref P 76.9% 16.0% 7.1%
x264 [info]: ref B 86.6% 9.5% 3.8%
x264 [info]: SSIM Mean Y:0.8846770
x264 [info]: PSNR Mean Y:34.463 U:42.216 V:43.612 Avg:35.896 Global:35.218
kb/s:1006.37

encoded 1236 frames, 8.26 fps, 1006.54 kb/s
desired video bitrate of this job: 1000 kbit/s - obtained video bitrate
(approximate): 1009 kbit/s




I have to make some graph in Excel about difference of those 3 profiles for every bitrate . Are these results OK for that or i didn't do proper coding? :confused: I also noticed that PSNR is not diplayed in [dB]. How do i get [dB] units?

[nm]

I have to make some graph in Excel about difference of those 3 profiles for every bitrate . Are these results OK for that or i didn't do proper coding? :confused:
There is something wrong with the high profile encode or the stats for it. Are you sure those are for the second pass and not the first?

People who are more familiar with MeGUI profiles can probably comment on your settings.

I also noticed that PSNR is not diplayed in [dB]. How do i get [dB] units?It is in dB scale.

[opti2k4]

You were right about high profile, i put 1st instead 2nd pass. Now its edited.

[Dark Shikari]

Your test is likely not a very good one because you're using such a low bitrate that your SSIM is below 0.9--something that nobody could imagine watching, given the atrocious quality such a number implies.

[CruNcher]

even SSIM of 0.7 can still look quiete acceptable

[Dark Shikari]

even SSIM of 0.7 can still look quiete acceptable
An SSIM of 0.7 looks like a Q10 JPEG image. Actually, probably worse.

[foxyshadis]

Although ssim .9 can look good, the ssim, psnr, and average quant all tell a story of lousy video. But:
bitrates 9,4.5,2,1,0.5,0.25Mbit
Obviously for this source 720p 1mbps is already too little and dropping it further is going to look much worse, but hey, it's for charting, not archiving.

The results you get look about right, there's not an enormous difference between the profiles on most movies. The differences will be content specific, animation makes better use of higher profiles.

[akupenguin]

even SSIM of 0.7 can still look quiete acceptable
How do you even get 0.7? I've never managed to make an encode below about 0.8, and that was CQ51.

[Dark Shikari]

How do you even get 0.7? I've never managed to make an encode below about 0.8, and that was CQ51.I've gotten near-zero SSIM by screwing up the encoder logic intentionally :p

[nm]

The results you get look about right, there's not an enormous difference between the profiles on most movies. The differences will be content specific, animation makes better use of higher profiles.
The noticeable PSNR and SSIM increase when going from HQ-Slowest to the much faster CE-Main is a bit surprising though, but I don't know about the particular settings used in those profiles to comment further.

[CruNcher]

aku this 0.7 result is not based on x264