Re: NHW Project development

--===============2339647970135647486==
Content-Type: multipart/alternative; 
	boundary="----=_Part_1790937_1542032448.1638482024920"

------=_Part_1790937_1542032448.1638482024920
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Raphael,



Some 15 years ago I implemented limited functionality Theora in our cameras=
 FPGA (it took me 6 month of hard labor), and then gave up - it is a very c=
rowded space and it is difficult to compete with more advanced codecs. And =
for our other work we anyway need almost raw image data, so we are using JP=
EG-based JP4 format (https://community.elphel.com/jp4) that we originally d=
eveloped for Google Books to compress raw Bayer mosaic images. Last few yea=
rs we work with the thermal images that need 16 bit, so far uncompressed TI=
FF files. We need lossless 2:1 compression (current recording bandwidth is =
about half of the image data rate (4 channels of 640x512x60fps, 16 bits), b=
ut it will be something simple (prediction by 4 neighbors), and no video. N=
ature of the thermal images is different from that of the conventional RGB =
- there is no scaling with exposure and environmental temperature causes sh=
ift of the pixel values, not scaling.



We did not post the latest results yet, the last blog is here - https://blo=
g.elphel.com/2021/07/calibration-of-the-lwir16-camera-prototype/



Andrey






---- On Thu, 02 Dec 2021 14:34:25 -0700 Raphael Canut <[email protected]> =
wrote ----


Hi Andrey,



Thank you for your answer.Actually, NHW is very low-power so I think it cou=
ld be geared toward any hardware.But actually I don't have hardware skills,=
 so NHW is not a real hardware project for now maybe? I agree with you that=
 it is extremely difficult (for me) to build a community around NHW, I'm ce=
rtainly very bad/underskilled at it.Any help is welcome!



Cheers,

Raphael



Le=C2=A0jeu. 2 d=C3=A9c. 2021 =C3=A0=C2=A021:13, Andrey Filippov <mailto:an=
[email protected]> a =C3=A9crit=C2=A0:

Raphael,



I do not believe it is possible to build a viable community around most har=
dware projects except super-universal based on very cheap/high-volume produ=
cts like Pi or Arduino. We see the commercial benefits of Free Software/Ope=
n Hardware as it adds value to the hardware as users are ready to pay extra=
 for the Freedom.



Andrey



---- On Thu, 02 Dec 2021 12:59:34 -0700 Raphael Canut <mailto:nhwcodec@gmai=
l.com> wrote ----



Hello,



I know there is no community around NHW, but I would have liked to know=20
 what you would like to see improved in the NHW codec.I know its results
  are not perfect for now compared to high-end AV1/VVC codecs for=20
example,  but could be ok giving its extremely fast speed.-However on=20
high  quality images, I notably like its good results because of its =20
neatness.-



Regarding things to improve, there is the problem of aliasing, I have =20
started some quick experiments as I mainly find that aliasing appears in
  NHW along "low-contrast edges", but I have also seen some very =20
impressive results of directional filters like the ones used in CDEF =20
filter in AV1, because they estimate the "strong edge direction" and =20
align all the pixels on that "strong edge direction"  which could =20
depixelize/remove aliasing.I have also started to implement a banding =20
detector (with its main analysis on the 128x128 wavelet DC image) in =20
order to remove banding, and it's only the beginning but I have =20
promising results.I have also started to improve the entropy coding =20
scheme of the wavelet DC part, for now I have a very fast processing =20
that gives another 1KB saving on that part in average, but I have an =20
error on some few images, I don't think I contradict the compression and
  information theory principles, I think it is more a corner case that I
  did not see for now and so that I don't handle, but it takes a lot of=20
 time... I'll also have to better handle the chroma channels and notably
  add chroma from luma technique.



Some people also tell me that the priority now is to adapt NHW to any =20
image size, but actually this is the boring task for me currently, and I
  am a little reluctant for now.



All that to let you know that I really like more to work on my codec, =20
and so I would like now to work full-time on it again, but there is =20
always the big problem of the salary at the end of the month.As the =20
industry doesn't show interest in NHW, I would like to ask sponsoring =20
for this project on the Internet.But some people told me that I failed =20
to create a community around NHW, and with just 27 followers on GitHub =20
and 8 followers on Twitter (that's very weak), I can not hope more than=20
 10dollars/per month of sponsoring/donation.



Do you think this solution/sponsoring is possible? Or on the contrary, =20
NHW will be just a spare time hobby, because there is nothing more to =20
expect out of it?



Any help/opinion is very welcome!



Cheers,

Raphael=20


_______________________________________________

theora mailing list=20

mailto:[email protected]=20

http://lists.xiph.org/mailman/listinfo/theora
------=_Part_1790937_1542032448.1638482024920
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"><html><head>=
<meta content=3D"text/html;charset=3DUTF-8" http-equiv=3D"Content-Type"></h=
ead><body ><div style=3D"font-family: Verdana, Arial, Helvetica, sans-serif=
; font-size: 10pt;"><div>Raphael,<br></div><div><br></div><div>Some 15 year=
s ago I implemented limited functionality Theora in our cameras FPGA (it to=
ok me 6 month of hard labor), and then gave up - it is a very crowded space=
 and it is difficult to compete with more advanced codecs. And for our othe=
r work we anyway need almost raw image data, so we are using JPEG-based JP4=
 format (<a target=3D"_blank" href=3D"https://community.elphel.com/jp4">htt=
ps://community.elphel.com/jp4</a>) that we originally developed for Google =
Books to compress raw Bayer mosaic images. Last few years we work with the =
thermal images that need 16 bit, so far uncompressed TIFF files. We need lo=
ssless 2:1 compression (current recording bandwidth is about half of the im=
age data rate (4 channels of 640x512x60fps, 16 bits), but it will be someth=
ing simple (prediction by 4 neighbors), and no video. Nature of the thermal=
 images is different from that of the conventional RGB - there is no scalin=
g with exposure and environmental temperature causes shift of the pixel val=
ues, not scaling.<br></div><div><br></div><div>We did not post the latest r=
esults yet, the last blog is here - <a target=3D"_blank" href=3D"https://bl=
og.elphel.com/2021/07/calibration-of-the-lwir16-camera-prototype/">https://=
blog.elphel.com/2021/07/calibration-of-the-lwir16-camera-prototype/</a><br>=
</div><div><br></div><div>Andrey<br></div><div><br></div><div class=3D"zmai=
l_extra_hr" style=3D"border-top: 1px solid rgb(204, 204, 204); height: 0px;=
 margin-top: 10px; margin-bottom: 10px; line-height: 0px;"><br></div><div c=
lass=3D"zmail_extra" data-zbluepencil-ignore=3D"true"><br><div id=3D"Zm-_Id=
_-Sgn1">---- On Thu, 02 Dec 2021 14:34:25 -0700 <b>Raphael Canut &lt;nhwcod=
[email protected]&gt;</b> wrote ----<br></div><br><blockquote style=3D"margin: 0=
px;"><div><div dir=3D"ltr"><div>Hi Andrey,<br></div><div><br></div><div>Tha=
nk you for your answer.Actually, NHW is very low-power so I think it could =
be geared toward any hardware.But actually I don't have hardware skills, so=
 NHW is not a real hardware project for now maybe? I agree with you that it=
 is extremely difficult (for me) to build a community around NHW, I'm certa=
inly very bad/underskilled at it.Any help is welcome!<br></div><div><br></d=
iv><div>Cheers,<br></div><div>Raphael<br></div></div><br><div class=3D"x_-1=
193805188gmail_quote"><div dir=3D"ltr" class=3D"x_-1193805188gmail_attr">Le=
&nbsp;jeu. 2 d=C3=A9c. 2021 =C3=A0&nbsp;21:13, Andrey Filippov &lt;<a href=
=3D"mailto:[email protected]" target=3D"_blank">[email protected]</a>&gt; a=
 =C3=A9crit&nbsp;:<br></div><blockquote class=3D"x_-1193805188gmail_quote" =
style=3D"margin :  0px 0px 0px 0.8ex; border-left :  1px solid rgb(204,204,=
204); padding-left :  1ex;"><u></u><div><div style=3D"font-family :  Verdan=
a, Arial, Helvetica, sans-serif; font-size :  10pt;"><div>Raphael,<br></div=
><div><br></div><div>I do not believe it is possible to build a viable comm=
unity around most hardware projects except super-universal based on very ch=
eap/high-volume products like Pi or Arduino. We see the commercial benefits=
 of Free Software/Open Hardware as it adds value to the hardware as users a=
re ready to pay extra for the Freedom.<br></div><div><br></div><div>Andrey<=
br></div><div><div><br></div><div id=3D"x_-1193805188gmail-m_81730718164344=
86294Zm-_Id_-Sgn1">---- On Thu, 02 Dec 2021 12:59:34 -0700 <b>Raphael Canut=
 &lt;<a href=3D"mailto:[email protected]" target=3D"_blank">nhwcodec@gmail=
.com</a>&gt;</b> wrote ----<br></div><div><br></div><blockquote style=3D"ma=
rgin :  0px;"><div><div dir=3D"ltr"><div>Hello,<br></div><div><br></div><di=
v>I know there is no community around NHW, but I would have liked to know=
=20
 what you would like to see improved in the NHW codec.I know its results
  are not perfect for now compared to high-end AV1/VVC codecs for=20
example,  but could be ok giving its extremely fast speed.-However on=20
high  quality images, I notably like its good results because of its =20
neatness.-<br></div><div><br></div><div>Regarding things to improve, there =
is the problem of aliasing, I have =20
started some quick experiments as I mainly find that aliasing appears in
  NHW along "low-contrast edges", but I have also seen some very =20
impressive results of directional filters like the ones used in CDEF =20
filter in AV1, because they estimate the "strong edge direction" and =20
align all the pixels on that "strong edge direction"  which could =20
depixelize/remove aliasing.I have also started to implement a banding =20
detector (with its main analysis on the 128x128 wavelet DC image) in =20
order to remove banding, and it's only the beginning but I have =20
promising results.I have also started to improve the entropy coding =20
scheme of the wavelet DC part, for now I have a very fast processing =20
that gives another 1KB saving on that part in average, but I have an =20
error on some few images, I don't think I contradict the compression and
  information theory principles, I think it is more a corner case that I
  did not see for now and so that I don't handle, but it takes a lot of=20
 time... I'll also have to better handle the chroma channels and notably
  add chroma from luma technique.<br></div><div><br></div><div>Some people =
also tell me that the priority now is to adapt NHW to any =20
image size, but actually this is the boring task for me currently, and I
  am a little reluctant for now.<br></div><div><br></div><div>All that to l=
et you know that I really like more to work on my codec, =20
and so I would like now to work full-time on it again, but there is =20
always the big problem of the salary at the end of the month.As the =20
industry doesn't show interest in NHW, I would like to ask sponsoring =20
for this project on the Internet.But some people told me that I failed =20
to create a community around NHW, and with just 27 followers on GitHub =20
and 8 followers on Twitter (that's very weak), I can not hope more than=20
 10dollars/per month of sponsoring/donation.<br></div><div><br></div><div>D=
o you think this solution/sponsoring is possible? Or on the contrary, =20
NHW will be just a spare time hobby, because there is nothing more to =20
expect out of it?<br></div><div><br></div><div>Any help/opinion is very wel=
come!<br></div><div><br></div><div>Cheers,<br></div><div>Raphael <br></div>=
</div><div>_______________________________________________<br></div><div>th=
eora mailing list <br></div><div><a href=3D"mailto:[email protected]" target=
=3D"_blank">[email protected]</a> <br></div><div><a href=3D"http://lists.xiph=
.org/mailman/listinfo/theora" target=3D"_blank">http://lists.xiph.org/mailm=
an/listinfo/theora</a> <br></div></div></blockquote></div><div><br></div></=
div><br></div></blockquote></div></div></blockquote></div><div><br></div></=
div><br></body></html>
------=_Part_1790937_1542032448.1638482024920--


--===============2339647970135647486==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
theora mailing list
[email protected]
http://lists.xiph.org/mailman/listinfo/theora

--===============2339647970135647486==--