High Definition with Elphel model 333 camera

[Wayne Morellini]

Originally Posted by Andrey Filippov

It is a big project to develop FPGA code for compressor. Theora took me some 2000 hours, Dirac will likely take more and I don't have such a window in my near future.

With Dirac they are doing their own, their is one being done at Opencores I think. Their is stuff afoot, and Dirac might not be the only one out there. So only a complete design need be integrated with support logic in the core.

What do you think of this 3D Wavelet, that is being used in Surveillance and other industries, is it any good?

[Wayne Morellini]

Here are some links, there is more going on I feel, then what is seen. I think, maybe, something usable for the next year or so:
http://dirac.sourceforge.net/
http://www.opencores.org/projects.cg...dirac/overview

Here are some others that might be of interest, but not FPGA:
http://en.wikipedia.org/wiki/Tarkin
http://en.wikipedia.org/wiki/Snow_(codec)
http://www.opencores.org/cvsweb.shtm...ssion_systems/ ;)


This for inspiration to beat, not such a great camera, but Bo's managed to get some good test footage out of it:
http://www.bophoto.com/HDV/video/

[Phil Stone]

Originally Posted by Wayne Morellini

This for inspiration to beat, not such a great camera, but Bo's managed to get some good test footage out of it:
http://www.bophoto.com/HDV/video/

I have an HD1 & its absolutely nothing like the elphel quality footage wise but its one fantastic camera to have for filming the kids & taking about in your pocket. Ive even used it as a helmet camera. But its image stabilization is awful so its useless for us jerky impatient dads out there!

If you look at the HD1 view example video of the hill & grass you can clearly see the terrible artifacts in the back ground with all the trees smudged in the distance with the grass on the distant hill being clear. The camera has a low bandwidth so you have to think about how complex the scene is if you want everything to be clear. In the clip it seems like its able to dedicate the bandwidth to the foreground grass which looks nice.

[Wayne Morellini]

Yes, but it still looks gorgeous so often in the right hands, that you can image how well an advanced Elphel could do? But, I would suggest look at the colour balance (when it is not oversaturated, see threads about turn saturation down, and off during playback) and response of the sensor under ideal conditions, this response does look pretty good compared to the Elphel footage I have seen.

[Phil Stone]

Originally Posted by Wayne Morellini

Yes, but it still looks gorgeous so often in the right hands, that you can image how well an advanced Elphel could do? But, I would suggest look at the colour balance (when it is not oversaturated, see threads about turn saturation down, and off during playback) and response of the sensor under ideal conditions, this response does look pretty good compared to the Elphel footage I have seen.

I will look for some raw elphel footage to post.

[Phil Stone]

http://www.tacx-video.com/raw-demo/R...3Raw-mjpeg.avi here is something I found on this laptop. Not ideal lighting (washed out sky) & its 4:3.

[Matteo Pozzi]

Thanks Phil for the link; if you find can you post some raw elphel footage in ogg theora ...exactly what you get from the camera? I want to experiment all the process...(from conversion to avi to color correction and editing with vegas)
best reguards Matteo

[Phil Stone]

Originally Posted by Matteo Pozzi

Thanks Phil for the link; if you find can you post some raw elphel footage in ogg theora ...exactly what you get from the camera? I want to experiment all the process...(from conversion to avi to color correction and editing with vegas)
best reguards Matteo

That is raw mjpeg (was in a ogg container) from the camera. ive not yet used the theora codec which I think is a form of Xvid. All I have done is change the container to avi.

Its 27.1fps 1280x1024 & very slightly out of focus.

here is the same clip compressed 10x so it looks a bit like HD1 video http://www.tacx-video.com/raw-demo/R...Compressed.avi

[Andrey Filippov]

Here is a screenshot of mostly routed 353 board - http://www.elphel.com/3fhlo/10353/353pcb.png

[Matteo Pozzi]

for Phil:
I thought you were shooting in ogg-theora and not ogg-mjpg ...can you tall me the maximum bitrate (more or less I know it is variable for jpg compression) that you can achieve and the maximum quality of jpg you can achieve from the camera with you portable setup!? (did you use a laptop right?)
for Andrey:
many thanks for what you are doing for us also for this picture ...I don't know anything about electronic scheme but it is really cool that you make us known your progress in your work.
In the final version of the 353 how did you intend to implement audio in the ogg container files? with a minijack or with a usb external box? is the usb 1.1 transfer rate able to record more than 2 track (stereo sound)?
best reguards
Matteo Pozzi

[Wayne Morellini]

Originally Posted by Phil Stone

That is raw MJpeg (was in a Ogg container) from the camera. ive not yet used the theora codec which I think is a form of Xvid. All I have done is change the container to avi.

Its 27.1fps 1280x1024 & very slightly out of focus.

here is the same clip compressed 10x so it looks a bit like HD1 video http://www.tacx-video.com/raw-demo/R...Compressed.avi

:)

To me the 100Mb/s MJpeg, restricts it to HVX200 quality, which uses the DV adaption of MJepg, at the most (the compressor runs at a lower rate I understand). Double HVX200 data rate is a good zone, and that is why I have always been interested in maximum Ogg Theora performance, because it si more efficient, to cram as much data is there as possible in 100mb/s, but I understand maximum rate of Ogg codec is around 30mb/s here. So next camera, I understand, will have much better through put, because ti si now probably faster, and Andrey mentioned improving motion vector etc, but looks like the model after 353. But to get equivalent to double the, per pixel, data rate of DVCPROHD, in 1920*1080 (not HVX200 1280*1080) is goign to require very good Ogg Theora. But then again, we have not tested off theora to prove quality.

I understand that Cinelerra Linux NLE supports Ogg Theora, found here:

Ogg Theora, at it's fullest is supposed to be an alternative to Mpeg2, but closer to performance of Mpeg4 I think I understand. It is an advanced form of the VP3 codec, which is now upto VP7, which is, supposed, to be better than H264. The camera to match and beat today is the HVX200 100mb/s DVCPROHD camera (much less than in 720p). But this old codec is being replaced by a frame based H264 Intra codec next year, that has indications of getting double DVCPROHD quality at 100Mb/s, with 10bit 4:2:2. So, there is possibility that a full Ogg Theora might have the possibility to stand against it, unless severe scene changes. There is also AVCHD, based on H264 Intra (intermediate frame with motion detection etc). It is possible that at 24mb/s it will be better than 100Mb/s DVCPROHD, unless severe scene changes like motion again.

The one big advantage of the present bandwidth, is that theoretically 100Mb/s can be sued for 720p, which the DVCPROHD cameras do not support, but how high was the compressor through put compared to this.

I understand that Ogg can do 4:2:2, and 4:4:4, but forgot to check for 10bits. Is 4:4:4/4:2:2 in this camera Andrey?

[Wayne Morellini]

Andrey,

Recently I have been looking up component to GigEthernet converters. I though, you could make one with a ADC front end, FPGA, and GigE conditioned back end (actually, trying to find some kind sole that might want to do it for us). Such a thing could be done DIY style for perhaps $100 with simple FPGA, and stream out to hard disk, or laptop in 10+ bits. Some simple, mild compression might be possible, or proper lossless compression with a bigger FPGA and more expense (like $500). You could then use it on a lot of HD cameras.

I then occurred to me, that you camera base could do this to hard disk (dropping the need for GigE) with only a mild adjustment to accept component, and record control signals from Firewire/lanc jack. All other camera and sensor control is done in the main body, so controls/battery/Elphel main board, and hard disk + component recording solution. If it can get at least 100Mb/s compression rate, it is useful cost effective equipment for some people. The only real thing needed on top of this, is to be able to input custom settings for the component readout sync and format configuration (even auto sync and remember) as some cameras output a unique component format and resolution (upto 14 bits) that can be processed into better image than normal component.

Does this sound doable?

[Matteo Pozzi]

Hi to all ....
http://photos1.blogger.com/blogger/3...gp0736_std.jpg
this is what I'm planning to do with the elphel 353 (this photo is from a silicon imaging 1920 camera)....by now I'm working to a
a project for a canon fd static adapter with nokon f3 type d screen and if it work properly I'll go for a mamiya 645 medium format adapter cause I wont to achieve 2.35 aspect ratio.
the lcd screen will be a touch screen lilliput 7" 800x480 (I've done a diy projector some time ago and after some internet research I think it is the best quality and cheaper panel you can own for about 200euro)
all the cable go to a notebook pc so:
camera--> ehernet to pc --> pc to lcd (vga cable) -->and from the lcd (usb cable) to pc for touchscreen mouse control
what do you think about?

[Wayne Morellini]

Nice, did you say it was a rig for a SI, what did you take the picture with.

[Matteo Pozzi]

The photo of the si camera is from the blog of David Newman that with it wins "Best Cinematography" award at the 48 hour film festival. http://cineform.blogspot.com/