I thought I'd do this shoutout for the impatiently curious hungering for the new video standard with an impending orgasm.
We compiled the latest HM reference software of the H265 standard and did some comparisons. I don't have screenshots at the moment because my friend accidentally wiped the machine state of the VM he compiled it on but we only compared it to x264 as we didn't really care about all the other lower-quality implementations and I doubt anyone else does either.
Settings on the x264 encodes were cranked to their max for best quality with CRF mode to reach the same filesize as the H265 output. Used high CRFs so the quality difference looks more obvious.
The H.265 video looked a little bit better, not that much but impressive considering the pre-standardization implement vs. x264 which has been fine-tuned for a decade to perfection.
H264 in its early days had worse quality than XviD so I think this is looking good.
+ Reply to Thread
Results 1 to 9 of 9
Can you recall the compression results? File size comparison to same video.
About 0.060 BPP. x264 at CRF24 and H265 at QP29, both output videos the same file size. 720p.
Really wish we at least did SSIM tests. For an estimation, if x264 scored 0.97600 H265 would probably get 0.98000. Better but only by a margin and its hard to tell at higher bitrates.
I did notice that H265 has much better noise and grain-preserving ability which'll be difficult for subjective metrics to properly rate. x264 tends to smear while H265 makes a good impression and not necessarily mathematically-accurate one, sort of like SBR for audio.
On clean videos it's harder to tell and I saw little difference on animated ones, likely because of the huge benefit of mb-tree that's yet to be implemented in any H265 encoder.
"both output videos the same file size" Hmmm...I thought that H265 was supposed to give smaller file sizes. That was supposed to be the reason for going to H265. Am I wrong there?
Comparisons are typically done at the same filesize or bitrate. You need to control all the variables in codec testing, and measure only 1 variable. This is scientific testing method.
If "video A" gives better visual quality, or higher metric quality than "video B" ....BUT the filesizes are different, then they use different bitrates. You cannot conclude anything...
Now if both use the same bitrate, but "video A" gives better visual quality , then that is evidence of something
Another method is to do sequential encodes of varying filesizes and plot the quality (e.g. ssim or psnr metric) at varying bitrates
i can't wait until divx releases their h265 profile so i can switch over to h265 for all my encodes.
A ip-tv company in china is first in world that use h.265 codec for streaming movies , currently there are 57 h.265 movies that you can watch ,player language is chinese but you can find your way through it.
Last edited by david55; 12th Apr 2013 at 08:09.
So will there be an x265 encoder, I hope?