[theora] NHW Project - new -l8 "very high compression" setting

[Raphael Canut]

Hello,

For those interested, I have added a -l8 "very high compression" setting to
the NHW Project.

I have added an experimental processing to remove details and preserve
neatness.I have also now a fast estimation of entropy of wavelet
coefficients and I use it accordingly to the image complexity but it is a
very first version... With these new processings, the entropy coding
schemes are less efficient, and so we could save for -l8 setting nearly 2KB
per compressed .nhw file in average, will have to improve them...

This new -l8 setting has so a good neatness to the detriment of details.It
is still experimental, but I find it competitive with x265 (HEVC).

Update at: http://nhwcodec.blogspot.com/

Any feedback is very welcome!

Cheers,
Raphael

[Raphael Canut]

Hello,

The -l8 "very high compression" setting I posted was not really good.Now I
have corrected and improved it, and now the -l8 quality setting is really
better, more precision and more neatness.

I think now it is a good setting, and I have started to compare it with
x265 and it can be now competitive.

More at: http://nhwcodec.blogspot.com/

Do not hesitate to test it and give me your feedback!

Many thanks!
Cheers,
Raphael

Post by Raphael Canut
Hello,
For those interested, I have added a -l8 "very high compression" setting
to the NHW Project.
I have added an experimental processing to remove details and preserve
neatness.I have also now a fast estimation of entropy of wavelet
coefficients and I use it accordingly to the image complexity but it is a
very first version... With these new processings, the entropy coding
schemes are less efficient, and so we could save for -l8 setting nearly 2KB
per compressed .nhw file in average, will have to improve them...
This new -l8 setting has so a good neatness to the detriment of details.It
is still experimental, but I find it competitive with x265 (HEVC).
Update at: http://nhwcodec.blogspot.com/
Any feedback is very welcome!
Cheers,
Raphael

[Raphael Canut]

Hello,

Did some of you find time to test the new -l8 "very high compression"
setting?

That's right that this new setting washes details out but on the other side
it keeps a good neatness of contours, edges.

Do you find it acceptable? If so, I could keep the new processing for -l9
setting (and certainly above).I have started to compare this new -l8
setting with x265 (HEVC), and on some many images (rather good quality) I
find it competitive.HEVC has a very very good precision, but the NHW
Project has more neatness, and on some images I prefer the results of the
NHW Project.

I forgot to tell that the new processings are ultra fast, and so the NHW
Project is always at least 50x faster to encode and at least 15x faster to
decode than x265! And it is still royalty-free!!!

As I found that it was a pity that the NHW Project could not be used in
mobile electronics, I decided one week ago to contact the Alliance for Open
Media and submit my codec... One week later, I still don't have answer, and
certainly that I will not have answer as AOMedia is not interested... (it's
hard to not have answer but it's like that...)

Again if some of you could test the last version of the NHW Project, I
would be very interested. Monty, could you find time to test it?

Many thanks!
Cheers,
Raphael

Post by Raphael Canut
Hello,
The -l8 "very high compression" setting I posted was not really good.Now I
have corrected and improved it, and now the -l8 quality setting is really
better, more precision and more neatness.
I think now it is a good setting, and I have started to compare it with
x265 and it can be now competitive.
More at: http://nhwcodec.blogspot.com/
Do not hesitate to test it and give me your feedback!
Many thanks!
Cheers,
Raphael

Post by Raphael Canut
Hello,
For those interested, I have added a -l8 "very high compression" setting
to the NHW Project.
I have added an experimental processing to remove details and preserve
neatness.I have also now a fast estimation of entropy of wavelet
coefficients and I use it accordingly to the image complexity but it is a
very first version... With these new processings, the entropy coding
schemes are less efficient, and so we could save for -l8 setting nearly 2KB
per compressed .nhw file in average, will have to improve them...
This new -l8 setting has so a good neatness to the detriment of
details.It is still experimental, but I find it competitive with x265
(HEVC).
Update at: http://nhwcodec.blogspot.com/
Any feedback is very welcome!
Cheers,
Raphael