[theora] NHW Project - speed comparison with x265

[Raphael Canut]

Hello,

I am very slowly working on the NHW Project.I recently made a speed
comparison with x265 and wanted to share it with you.

On my processor Intel Core i5-6400, in average the NHW encoder (totally
unoptimized) is x10 times faster to encode than x265 (png decoding time
removed), in average 30ms vs 300ms for a 512x512 24bit color image, and in
average the NHW decoder (totally unoptimized) is x3 times faster to decode
than x265 (output to .ppm file), in average 10ms vs 30ms.

As a reminder, x265 is ultra optimized, so with good C optimization, SIMD
optimization and multithreading, the NHW codec will be x40 times faster to
encode and x12 times faster to decode than x265 or even more, so this
confirms that the NHW Project could be a good candidate for the mobile
device market.

At mid-compression, quality-wise I find (visual tests) that the NHW codec
and x265 are on par, sometimes more neatness is better, sometimes more
precision is better, and a lot of improvements remain to be done for the
NHW codec.

If there are engineers from mobile device companies on this forum, could
you consider the NHW Project or is it too late it will be AOM AV1 or HEVC
despite their high power consumption?

Monty, would it be possible that the NHW Project has a Xiph.org page to
encourage and boost its development?

Any answer welcome!

Cheers,
Raphael

[fboehm]

Hi Raphael,

since quite a while I'm reading your updates on your NHW project with
great interest.

I have to admit I'm no video codec developer. My interest in video
codecs is pretty simple. Several years ago I was working on a
(non-commercial) video streaming project that I wanted to base on an
open-source video codec. Those I was constantly scouting for open source
video codec innovations.

I appreciate your enthusiastic work on such a complex project like
writing a video codec. As you already realized it's similarly
challenging to attract other people to join or use a new video codec.

What I unfortunately can say for sure is, that it's almost impossible to
attract mobile/semiconductor industry to embed such a codec into their
chipsets. Instead you have to find some other (maybe niche) use-case
where your codec works well and can blossom. As an example, maybe some
sort of decentralized/peer-to-peer Internet enthusiasts would enojy
using your codecto develop some sort of open-source alternative to Skype.

To make a long story short: to gain more publicity you really have to
create a website with visual presentations of your codecs quality. And
please, please, please present your results in an appealing way. In the
base case use a javascript-image-slider effect. Example:

https://codyhouse.co/demo/image-comparison-slider/index.html

And of course it would be great if Xiph could host your demo website.

All the best,
Franz

Post by Raphael Canut
Hello,
I am very slowly working on the NHW Project.I recently made a speed
comparison with x265 and wanted to share it with you.
On my processor Intel Core i5-6400, in average the NHW encoder (totally
unoptimized) is x10 times faster to encode than x265 (png decoding time
removed), in average 30ms vs 300ms for a 512x512 24bit color image, and
in average the NHW decoder (totally unoptimized) is x3 times faster to
decode than x265 (output to .ppm file), in average 10ms vs 30ms.
As a reminder, x265 is ultra optimized, so with good C optimization,
SIMD optimization and multithreading, the NHW codec will be x40 times
faster to encode and x12 times faster to decode than x265 or even more,
so this confirms that the NHW Project could be a good candidate for the
mobile device market.
At mid-compression, quality-wise I find (visual tests) that the NHW
codec and x265 are on par, sometimes more neatness is better, sometimes
more precision is better, and a lot of improvements remain to be done
for the NHW codec.
If there are engineers from mobile device companies on this forum, could
you consider the NHW Project or is it too late it will be AOM AV1 or
HEVC despite their high power consumption?
Monty, would it be possible that the NHW Project has a Xiph.org page to
encourage and boost its development?
Any answer welcome!
Cheers,
Raphael
_______________________________________________
theora mailing list
http://lists.xiph.org/mailman/listinfo/theora

[Raphael Canut]

---------- Forwarded message ----------
From: Raphael Canut <***@gmail.com>
Date: 2017-11-30 14:22 GMT+01:00
Subject: Re: [theora] NHW Project - speed comparison with x265
To: fboehm <***@aon.at>


Hi Franz,

Many thanks for your answer!!!

Post by fboehm
I appreciate your enthusiastic work on such a complex project like

writing a video codec

Yes, for now the NHW codec is more an image codec actually, would need to
add inter frame mecanism (motion estimation, motion compensation) to the
codec to become a real video codec.The NHW Project is more a
proof-of-concept in its currents state to validate for example the wavelet
transform, the multistage residual coding, the compression schemes, and
other processings, that are new and royalty-free.Do you think I can have a
freelance contract with minimum salary to do this, because completing and
finishing the codec will require me full-time work.

Post by fboehm
Instead you have to find some other (maybe niche) use-case where your

codec works well and can blossom. As an example, maybe some sort of
decentralized/peer-to-peer Internet enthusiasts > would enojy using your
codecto develop some sort of open-source alternative to Skype.

Would be just great to find a use-case for the NHW codec! If you and/or
other Xiph community members would know such potential use-case,
niche-case, do not hesitate to let me know!

Post by fboehm
To make a long story short: to gain more publicity you really have to

create a website with visual presentations of your codecs quality. And
please, please, please present your results in an

Post by fboehm
appealing way. In the base case use a javascript-image-slider effect.

Yes the NHW Project lacks a good website with good visuals.But I won't be
able to do this website alone...

As you suggest, maybe Xiph.org wants to host and help with the NHW Project
website? -Would be just great!-

Again thank you very much for your kind help, it is much appreciated.

Kind regards,
Raphael

Post by fboehm
Hi Raphael,
since quite a while I'm reading your updates on your NHW project with
great interest.
I have to admit I'm no video codec developer. My interest in video codecs
is pretty simple. Several years ago I was working on a (non-commercial)
video streaming project that I wanted to base on an open-source video
codec. Those I was constantly scouting for open source video codec
innovations.
I appreciate your enthusiastic work on such a complex project like writing
a video codec. As you already realized it's similarly challenging to
attract other people to join or use a new video codec.
What I unfortunately can say for sure is, that it's almost impossible to
attract mobile/semiconductor industry to embed such a codec into their
chipsets. Instead you have to find some other (maybe niche) use-case where
your codec works well and can blossom. As an example, maybe some sort of
decentralized/peer-to-peer Internet enthusiasts would enojy using your
codecto develop some sort of open-source alternative to Skype.
To make a long story short: to gain more publicity you really have to
create a website with visual presentations of your codecs quality. And
please, please, please present your results in an appealing way. In the
https://codyhouse.co/demo/image-comparison-slider/index.html
And of course it would be great if Xiph could host your demo website.
All the best,
Franz

Post by Raphael Canut
Hello,
I am very slowly working on the NHW Project.I recently made a speed
comparison with x265 and wanted to share it with you.
On my processor Intel Core i5-6400, in average the NHW encoder (totally
unoptimized) is x10 times faster to encode than x265 (png decoding time
removed), in average 30ms vs 300ms for a 512x512 24bit color image, and in
average the NHW decoder (totally unoptimized) is x3 times faster to decode
than x265 (output to .ppm file), in average 10ms vs 30ms.
As a reminder, x265 is ultra optimized, so with good C optimization, SIMD
optimization and multithreading, the NHW codec will be x40 times faster to
encode and x12 times faster to decode than x265 or even more, so this
confirms that the NHW Project could be a good candidate for the mobile
device market.
At mid-compression, quality-wise I find (visual tests) that the NHW codec
and x265 are on par, sometimes more neatness is better, sometimes more
precision is better, and a lot of improvements remain to be done for the
NHW codec.
If there are engineers from mobile device companies on this forum, could
you consider the NHW Project or is it too late it will be AOM AV1 or HEVC
despite their high power consumption?
Monty, would it be possible that the NHW Project has a Xiph.org page to
encourage and boost its development?
Any answer welcome!
Cheers,
Raphael
_______________________________________________
theora mailing list
http://lists.xiph.org/mailman/listinfo/theora

_______________________________________________
theora mailing list
http://lists.xiph.org/mailman/listinfo/theora