[theora] NHW Project - Image comparison

Raphael Canut nhwcodec at gmail.com
Sat Aug 4 16:20:10 UTC 2018


Hello,

Could some of you take a look at the image comparison between the NHW
Project and x265 (HEVC) at high compression, available here:
http://nhwcodec.blogspot.com/ .

I would be very very interested in your opinion, and it would help me to
perfect the NHW Project.For me the NHW Project has more neatness, and x265
has more precision.

Else, I was a little lazy these past months, as I didn't work on the
codec.It mainly remains very high compression, and my idea would be, a
little like HEVC, to strongly remove the details inside the contours of
image, and maintains sharp and clean edges, contours.What would you think
of it? Would you have some algorithm ideas?

Else it is very hard to find some interest for the NHW Project, few days
ago I posted on the IETF video-codec mailing list but couldn't have any
answer, interest...

I still think that the NHW Project could be a good codec for mobile,
embedded devices as it has a very low power and ressource consumption.

For remind and for those who didn't read my previous post, here are salient
features and speed timings between the NHW Project and x265 codecs:

For encoding, the NHW Project (with all the expensive processings turned
on: pre-processing, feedback correction, residual coding, dithering,...)
takes 30ms to encode a 512x512 24bit bitmap image, and the x265 codec takes
300ms to encode that same ppm image.

For decoding, the NHW Project (with all the expensive processing turned on:
post-processing, UV comp sharpening,...) takes 10ms to decode the image,
and the x265 codec takes 30ms to decode that same ppm image.

As a remark, the x265 codec is highly highly optimized, notably SIMD and
multithreading, and the NHW Project has absolutely no optimization (pure C
unoptimized code).So with the same optimizations as x265 codec (x5), the
NHW Project is at least 50x faster to encode and at least 15x faster to
decode than x265 codec (HEVC)!!!

Any opinion, feedback are very welcome!

Many thanks!
Cheers,
Raphael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xiph.org/pipermail/theora/attachments/20180804/45ade49c/attachment.html>


More information about the theora mailing list