3 channel 36 bit 1280 X 720 low $ camera - Viper?

Forrest Schultz · February 17th, 2006, 05:37 PM

hey, how bout this also

http://www.visioncom.co.il/cameras/M...r_PL-A654.html

i dont know the price at all though. Its a firewire camera, uses a 1/2" cmos sensor. uses c-mount. and is uncompressed video. it can do 1024 x 768 at 23 fps. how much is this imaging module??? comes with software. anyone know the price? if its not much more, id go with this one.

Joel Kreisel · February 17th, 2006, 05:41 PM

If you guys haven't found them before, Edmund Optics has quite a few cameras like this (and with prices too!) with all manner of features, frame rates and connection types. I've already got a camera so I'm not particularly looking for another, but maybe eventually I'll replace the one I've got with a custom-built one...

Wayne Morellini · February 18th, 2006, 08:00 AM

As far as I can remember I think there is a syncing mechanism between the cameras. Yes. Forrest's system can work, the Olympus 8Mp ENG camera uses a prism to do the same thing. Thermal expansion and physical bumps can put them out of alignment, prism is better. Getting them to line up pixel perfect side by side is best, but not so easy, unless you make special cross hair chart for alignment. After you buy 4, you might as well experimented with www.Elphel.com HD compressed security camera for less than $800.

Pixel shift
-----------------------------------------------
How pixel shift works is that you get 640*480 VGA images planes of red green and blue. The images are aligned so that (for instance) what the green plain is recording is offset 1/3 rd a pixel down and left from the red plain, and what the blue plane is offset 1/3 rd a pixel down and left from the green plain. You do this by shifting the sensors. It produces three times more virtual pixels in both the horizontal and vertical directions (I.E. 9 times more pixels). Unlike Bayer there is no missing data to guess (interpolation) as all layers cover everything except for a few lines on the edges of the screen, where there are only one or two planes working (use interpolation techniques there).

So instead of the red green and blue plains being exactly aligned there overlap is offset. Because they are overlapping they record (this is very similar to how a 300dpi ink jet printer can get 1200dpi edges) a slightly different location in the images. But what happens in the overlap is that now each pixel on the different plains sticks out slightly from the last plane, giving a fine detailed artificial edge. Confusing, I am not really upto explaining it. But because this is done in horizontal and vertical directions the horizontal edges split across the vertical edges, making a fine series of sub-pixels. So unprocessed, this will give you a HD image. But on a resolution chart you will find that it seems to blur pixel edges. But because light and colour across objects follow predictable patterns in nature, you can make the software identify details and edges, and scale the resolution up and remove or add to the pixels to move the edges and details into their correct resolution.

Pixel shifting is done a fair bit on cameras, but usually only the green plane is 50% shifted. Pro SD ENG cameras might move the green plain to cover the space in between the pixel that produces the screen door effect etc. The Canon XL1 used it, that is why it got better low light performance and bigger latitude in it's day. The Panasonic AGX100 uses it, that is how the Andromeda conversion gets HD from it, by using the pixel shift to do it. The Sony Z1, Canon (don't know about the JVC) and I think the HVX200 use it. The cheap 3chip Panasonic have been using it. I think it was the GS120, people were not so impressed with the resolution charts (blurring) from the pixel shift. I think the GS120/65 use 1/3rd pixel shift.

What I am saying is that processed the, results probably won't be any worse than debayered or 4:2:0 video. Unprocessed, maybe de-Bayer can look better, but I think it might turn out better than 4:2:0.

I'm not too on the ball at the moment, if anybody can explain it simpler, please do.

-----------------------------------------------

About the two cameras taking alternative frame thing, somebody was looking at dong that with still cameras in recent months. Search for all threads I have posted to and you'll probably find it. Might even be a Red thread.

If Ronald posted it, it is worth looking up.

I would like to suggest that you start up a new thread on this topic and post a link here, as this is a thread on the extinct Sumix camera ;)

Forrest Schultz · February 18th, 2006, 08:49 AM

I got ya, thanks wayne! ive got a couple of choices to think about right now. so imm gonna have to settle on one.

Oscar Spierenburg · February 18th, 2006, 07:47 PM

Does anyone know more about these Elphel camera's?
These sample images look very impressive.

Daniel Rudd · February 18th, 2006, 09:50 PM

wow. looks promising. to a completely untrained eye.
I'm eager to learn more. Did you e-mail him Oscar?

I noticed an SD resolution at 60fps.

I wouldn't mind combining 4 of these focusing on four segements of a ground glass, synced up to a fast laptop yielding uncompressed 60p HD

Wayne Morellini · February 19th, 2006, 05:18 AM

I have been emailing him over converting his cameras to cinema use, I actual have mail waiting from him. He is content with the security camera business, though there is much opportunity, I can appreciate that it must be very hard to move in more than one direction when your business is not large. I've been in occasional contact with him since 2004.

You would be in the same position as with a Sumix, or a Silicon imaging camera (except you get a lot more for your money and the software development stuff I think is free, not many hundreds of dollars) and no capture card needed for many computers. It probably would be best to contact the Ogg people to find out about capture and editing with the Ogg codec he uses. The Linux open source community networking with open source programmers, particularly with ones interested in things like the Cinelerra editor/capture program, might find somebody interested in helping. If you look on the source forge for Elphel, you will see somebody offering to party programmers to develop a application for industrial inspection use (well that's what I think they are doing). So Elphel community, Cinelerra, Ogg open-source community would probably be the best place to try.

Another word, without expanding the FPGA to record directly to disk, the next best option to by pass a computer, would be an embedded board that controls a external caddy hard disk and passes the data directly to it (automated pref no processor intervention all) at he same time copying the information to it's memory for a preview function to run off an external LCD, also controlling buttons and menu system and passing commands to the camera. All easier said than done. Small Arm/Mips/micro-controller boards may have all the integrated systems needed to do this, including video output.

Oscar Spierenburg · February 24th, 2006, 05:59 AM

I see a member called Larry Liang posted about their camera on this board:

<<<<<Elphel's camera use the micron's 1.3 Mega and 3 Mega sensor. It can output Motion-Jpeg stream with a speed of 22 fps@1280*1024, 31fps@1280*720.

With Elphel's camera, it is much easier for storage, because the video has been compressed before output, but hard to display: you need additional CPU capacity of decoding the JPEG picture. Generally, To display the video with full speed and resolution, a 3G P4.CPU is needed.
>>>>>>

Does anyone know how good or bad the compression is? How about using such a camera (or other) and using a separate viewfinder to eliminate preview during capture.

Wayne Morellini · February 26th, 2006, 05:38 AM

He has also moved to Ogg Theodore in his latest camera, I don't know anything about the codec, but if you look up sourceforge you can find information, and the website for Ogg and more information. On the Elphel site (also look at their wiki and source forge, they have sample footage).

Régine Weinberg · February 27th, 2006, 08:29 AM

http://www.imperx.com/machine_vision...ion/index.html

have a look there may not be cheapo but
1920 x 1080 pixels @ 33 fps
8/10/12 bit data
Gig E (Gigabit Ethernet

Wayne Morellini · February 27th, 2006, 11:37 PM

Ronald,

Does it have a proper video capture application? Is there any reasonably priced camera that has one?

Oscar Spierenburg · March 1st, 2006, 07:14 PM

Have you guys noticed that most of the info on the Elphel 333 model is located on this site: http://sourceforge.net/projects/elphel
They have an interesting forum, someone is actually using it to film footage for his bike-race simulating software (for home-trainer bikes). It means he films the road from a car or something, I guess those are long takes.
I downloaded all the info incl. pricing of the 333 model here:
http://sourceforge.net/project/showf...ease_id=391625 (elphel_docs-2006-02-07.tar.bz2)

More about the sensor site is here: http://www.micron.com/products/imagi...s/MT9T001.html

Is it me (who knows too little about all this) or is this really promising?
What do you people think would be the minimal system requirements to capture (without preview) from this camera, say 22Fps at 720p or more?

Forrest Schultz · March 1st, 2006, 08:09 PM

I think this camera is very promising, when i get the cash, and the know-how about making a portable recording solution. im going to buy one and make it into a cinema camera.

Wayne Morellini · March 2nd, 2006, 07:27 AM

Actually, this camera is probably one of the easier ones for recording, the stream is somewhere below 100Mb/s, so drives that can sustain this and processing bus requirements are easier. The previews is another thing though, that I do not know about. I suspect an embedded board that passes the stream to an Ethernet Hard drive caddy, and does a preview, as a possibility.

Oscar Spierenburg · March 2nd, 2006, 09:21 AM

About previewing the image.. My idea would be to integrate such a camera in a 35mm adapter (possibly a good microwax GG) and use a two way mirror or just a piece of anti reflective glass, between the GG and the camera, to either make an optical viewfinder on top, or a (cheap)small B/W CMOS camera and use an LCD to view the image on the mirror. I got the idea from Dan Diaconu, so it should be worth a shot.

Focusing this way won't be accurate enough, I'm sure, but for me it'll be fine to measure the right distance to the subject. (maybe with a laser measure or something)

I wrote an email to the Belgian guy who uses the camera to capture cycling races to his laptop. I'll come back to this when I know more.

February 18th, 2006, 09:50 PM	#306
Daniel Rudd Major Player Join Date: Aug 2005 Location: North Muskegon, MI Posts: 213	wow. looks promising. to a completely untrained eye. I'm eager to learn more. Did you e-mail him Oscar? I noticed an SD resolution at 60fps. I wouldn't mind combining 4 of these focusing on four segements of a ground glass, synced up to a fast laptop yielding uncompressed 60p HD __________________ Daniel Rudd Digital Storyteller (Sony HDV, Aspect HD) Soundtrack Creation & Royalty Free Music Production www.stock20.com

February 27th, 2006, 08:29 AM	#310
Régine Weinberg Major Player Join Date: Jan 2004 Location: Bordeaux, going to Bangkok, 2011 Posts: 232	1920x1080 33fps GigE http://www.imperx.com/machine_vision...ion/index.html have a look there may not be cheapo but 1920 x 1080 pixels @ 33 fps 8/10/12 bit data Gig E (Gigabit Ethernet

February 17th, 2006, 05:37 PM	#301
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	hey, how bout this also http://www.visioncom.co.il/cameras/M...r_PL-A654.html i dont know the price at all though. Its a firewire camera, uses a 1/2" cmos sensor. uses c-mount. and is uncompressed video. it can do 1024 x 768 at 23 fps. how much is this imaging module??? comes with software. anyone know the price? if its not much more, id go with this one.

February 17th, 2006, 05:41 PM	#302
Joel Kreisel Regular Crew Join Date: Jan 2006 Location: Minneapolis, MN, USA Posts: 44	If you guys haven't found them before, Edmund Optics has quite a few cameras like this (and with prices too!) with all manner of features, frame rates and connection types. I've already got a camera so I'm not particularly looking for another, but maybe eventually I'll replace the one I've got with a custom-built one...

February 18th, 2006, 08:00 AM	#303
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	As far as I can remember I think there is a syncing mechanism between the cameras. Yes. Forrest's system can work, the Olympus 8Mp ENG camera uses a prism to do the same thing. Thermal expansion and physical bumps can put them out of alignment, prism is better. Getting them to line up pixel perfect side by side is best, but not so easy, unless you make special cross hair chart for alignment. After you buy 4, you might as well experimented with www.Elphel.com HD compressed security camera for less than $800. Pixel shift ----------------------------------------------- How pixel shift works is that you get 640*480 VGA images planes of red green and blue. The images are aligned so that (for instance) what the green plain is recording is offset 1/3 rd a pixel down and left from the red plain, and what the blue plane is offset 1/3 rd a pixel down and left from the green plain. You do this by shifting the sensors. It produces three times more virtual pixels in both the horizontal and vertical directions (I.E. 9 times more pixels). Unlike Bayer there is no missing data to guess (interpolation) as all layers cover everything except for a few lines on the edges of the screen, where there are only one or two planes working (use interpolation techniques there). So instead of the red green and blue plains being exactly aligned there overlap is offset. Because they are overlapping they record (this is very similar to how a 300dpi ink jet printer can get 1200dpi edges) a slightly different location in the images. But what happens in the overlap is that now each pixel on the different plains sticks out slightly from the last plane, giving a fine detailed artificial edge. Confusing, I am not really upto explaining it. But because this is done in horizontal and vertical directions the horizontal edges split across the vertical edges, making a fine series of sub-pixels. So unprocessed, this will give you a HD image. But on a resolution chart you will find that it seems to blur pixel edges. But because light and colour across objects follow predictable patterns in nature, you can make the software identify details and edges, and scale the resolution up and remove or add to the pixels to move the edges and details into their correct resolution. Pixel shifting is done a fair bit on cameras, but usually only the green plane is 50% shifted. Pro SD ENG cameras might move the green plain to cover the space in between the pixel that produces the screen door effect etc. The Canon XL1 used it, that is why it got better low light performance and bigger latitude in it's day. The Panasonic AGX100 uses it, that is how the Andromeda conversion gets HD from it, by using the pixel shift to do it. The Sony Z1, Canon (don't know about the JVC) and I think the HVX200 use it. The cheap 3chip Panasonic have been using it. I think it was the GS120, people were not so impressed with the resolution charts (blurring) from the pixel shift. I think the GS120/65 use 1/3rd pixel shift. What I am saying is that processed the, results probably won't be any worse than debayered or 4:2:0 video. Unprocessed, maybe de-Bayer can look better, but I think it might turn out better than 4:2:0. I'm not too on the ball at the moment, if anybody can explain it simpler, please do. ----------------------------------------------- About the two cameras taking alternative frame thing, somebody was looking at dong that with still cameras in recent months. Search for all threads I have posted to and you'll probably find it. Might even be a Red thread. If Ronald posted it, it is worth looking up. I would like to suggest that you start up a new thread on this topic and post a link here, as this is a thread on the extinct Sumix camera ;)

February 18th, 2006, 08:49 AM	#304
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	I got ya, thanks wayne! ive got a couple of choices to think about right now. so imm gonna have to settle on one.

February 18th, 2006, 07:47 PM	#305
Oscar Spierenburg Major Player Join Date: Jan 2005 Location: (The Netherlands - Belgium) Posts: 735	Does anyone know more about these Elphel camera's? These sample images look very impressive.

February 19th, 2006, 05:18 AM	#307
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I have been emailing him over converting his cameras to cinema use, I actual have mail waiting from him. He is content with the security camera business, though there is much opportunity, I can appreciate that it must be very hard to move in more than one direction when your business is not large. I've been in occasional contact with him since 2004. You would be in the same position as with a Sumix, or a Silicon imaging camera (except you get a lot more for your money and the software development stuff I think is free, not many hundreds of dollars) and no capture card needed for many computers. It probably would be best to contact the Ogg people to find out about capture and editing with the Ogg codec he uses. The Linux open source community networking with open source programmers, particularly with ones interested in things like the Cinelerra editor/capture program, might find somebody interested in helping. If you look on the source forge for Elphel, you will see somebody offering to party programmers to develop a application for industrial inspection use (well that's what I think they are doing). So Elphel community, Cinelerra, Ogg open-source community would probably be the best place to try. Another word, without expanding the FPGA to record directly to disk, the next best option to by pass a computer, would be an embedded board that controls a external caddy hard disk and passes the data directly to it (automated pref no processor intervention all) at he same time copying the information to it's memory for a preview function to run off an external LCD, also controlling buttons and menu system and passing commands to the camera. All easier said than done. Small Arm/Mips/micro-controller boards may have all the integrated systems needed to do this, including video output.

February 24th, 2006, 05:59 AM	#308
Oscar Spierenburg Major Player Join Date: Jan 2005 Location: (The Netherlands - Belgium) Posts: 735	I see a member called Larry Liang posted about their camera on this board: <<<<<Elphel's camera use the micron's 1.3 Mega and 3 Mega sensor. It can output Motion-Jpeg stream with a speed of 22 fps@12801024, 31fps@1280720. With Elphel's camera, it is much easier for storage, because the video has been compressed before output, but hard to display: you need additional CPU capacity of decoding the JPEG picture. Generally, To display the video with full speed and resolution, a 3G P4.CPU is needed. >>>>>> Does anyone know how good or bad the compression is? How about using such a camera (or other) and using a separate viewfinder to eliminate preview during capture.

February 26th, 2006, 05:38 AM	#309
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	He has also moved to Ogg Theodore in his latest camera, I don't know anything about the codec, but if you look up sourceforge you can find information, and the website for Ogg and more information. On the Elphel site (also look at their wiki and source forge, they have sample footage).

February 27th, 2006, 11:37 PM	#311
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Ronald, Does it have a proper video capture application? Is there any reasonably priced camera that has one?

March 1st, 2006, 07:14 PM	#312
Oscar Spierenburg Major Player Join Date: Jan 2005 Location: (The Netherlands - Belgium) Posts: 735	Have you guys noticed that most of the info on the Elphel 333 model is located on this site: http://sourceforge.net/projects/elphel They have an interesting forum, someone is actually using it to film footage for his bike-race simulating software (for home-trainer bikes). It means he films the road from a car or something, I guess those are long takes. I downloaded all the info incl. pricing of the 333 model here: http://sourceforge.net/project/showf...ease_id=391625 (elphel_docs-2006-02-07.tar.bz2) More about the sensor site is here: http://www.micron.com/products/imagi...s/MT9T001.html Is it me (who knows too little about all this) or is this really promising? What do you people think would be the minimal system requirements to capture (without preview) from this camera, say 22Fps at 720p or more?

March 1st, 2006, 08:09 PM	#313
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	I think this camera is very promising, when i get the cash, and the know-how about making a portable recording solution. im going to buy one and make it into a cinema camera.

March 2nd, 2006, 07:27 AM	#314
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Actually, this camera is probably one of the easier ones for recording, the stream is somewhere below 100Mb/s, so drives that can sustain this and processing bus requirements are easier. The previews is another thing though, that I do not know about. I suspect an embedded board that passes the stream to an Ethernet Hard drive caddy, and does a preview, as a possibility.

March 2nd, 2006, 09:21 AM	#315
Oscar Spierenburg Major Player Join Date: Jan 2005 Location: (The Netherlands - Belgium) Posts: 735	About previewing the image.. My idea would be to integrate such a camera in a 35mm adapter (possibly a good microwax GG) and use a two way mirror or just a piece of anti reflective glass, between the GG and the camera, to either make an optical viewfinder on top, or a (cheap)small B/W CMOS camera and use an LCD to view the image on the mirror. I got the idea from Dan Diaconu, so it should be worth a shot. Focusing this way won't be accurate enough, I'm sure, but for me it'll be fine to measure the right distance to the subject. (maybe with a laser measure or something) I wrote an email to the Belgian guy who uses the camera to capture cycling races to his laptop. I'll come back to this when I know more.