Interest in open-source capture software for Sumix M73

Cesar Rubio · December 30th, 2006, 03:38 PM

Igor:

With Stream Pix you DON'T loose any video quality recording in RAW mode, on the contrary you can SAVE all the information that the sensor CCD or CMOS is able to offer, and you can get the BEST video quality in post later on. With SP you also can record audio in sync with your video...give it a try, it's good.

Cesar Rubio.

John Wyatt · January 1st, 2007, 11:37 AM

Igor -- the M73 (like many other box cameras) does not record sound in any way, consequently there is no requirement for the capture software to record sound. My own plan for sound is to record it with a DV camcorder (16-bit stereo) running at the same time -- marry up sound with low res proxy later for off-line editing. This sort of set-up (and the need for a sound synching method) may not be suitable for your project if you need high levels of mobility and informal shooting with long takes.

Cesar -- it's great to see someone championing 3D! You mentioned StreamPix capture software; I looked at this about two years ago but was put off by the cost. It seemed too expensive for no budget projects like mine; has it come down in price lately?

Rob -- can you tell me if you have been able to capture 10-bit images from the M73 with your software? That is an area which really interests me, because it takes the camera even further away from an HDV camcorder.

Happy New Year, everyone.

Rob Scott · January 2nd, 2007, 10:53 AM

Quote:

Originally Posted by John Wyatt

Rob -- can you tell me if you have been able to capture 10-bit images from the M73 with your software? That is an area which really interests me, because it takes the camera even further away from an HDV camcorder.

I believe so, but I've take a few weeks off the project, so I'll have to go back to my notes and verify that. The biggest problem with 10-bit (vs. 8-bit) is that it cuts the bandwidth in half, making the the rolling shutter artifacts very pronounced. From what I can tell, to reduce artifacts enough with the M73 you'd need to stick with 8 bits or use 10 bits with a mechanical shutter.

Rob Scott · January 2nd, 2007, 11:04 AM

Quote:

Originally Posted by John Wyatt

My own plan for sound is to record it with a DV camcorder (16-bit stereo) running at the same time -- marry up sound with low res proxy later for off-line editing. This sort of set-up (and the need for a sound synching method) may not be suitable for your project if you need high levels of mobility and informal shooting with long takes.

Just FYI, at some point I am going to try capturing simultaneous audio in ObscuraCap.

Rob Scott · January 2nd, 2007, 11:07 AM

Quote:

Originally Posted by Igor Babic

Rob,
I am intrested in your project. I have looking at SMX73 for an indie film project where this camera will be attached to helmet for first person view.

Sounds interesting!

Quote:

Originally Posted by Igor Babic

VirtualDub has excellent Capture utility and is famous for precise audio sinc to video. It has lots of real time info during capture: File size, Frames captured, CPU usage, Data rate, Compression ratio, Frame dropped, VU meter, Histogram, realtime preview.

I believe the biggest issues are (1) limited to 8 bit depth and (2) we'd have to do realtime Bayer filtering.

John Wyatt · January 2nd, 2007, 08:40 PM

Rob -- thanks for the latest info: "To reduce artifacts enough with the M73 you'd need to stick with 8 bits" -- shame, 10-bit would have been the cherry on top; I should be more grateful though for what we do have with 8-bit! Rob, do you have any recording durations yet; I was wondering how much longer shooting duration might be compared to the supplied software? Also interested in your opinion of the deBayer you're using compared to the Laplacian one with the supplied software.

"I am going to try capturing simultaneous audio in ObscuraCap" -- sounds good Rob; perceived wisdom is it's best not to do anything else with the computer since it's probably maxed out taking in the video (and you don't want any background interrupts to that), though if it can be done (instead of using a separate device), that's obviously a good thing for portability. Thanks once again for your effort on this...

Rob Scott · January 2nd, 2007, 09:23 PM

Quote:

Originally Posted by John Wyatt

Rob, do you have any recording durations yet; I was wondering how much longer shooting duration might be compared to the supplied software?

With the M73 in 1024x576 (binned) mode, I am expecting to be able to be able to capture indefinitely (or until the drive is full).

Quote:

perceived wisdom is it's best not to do anything else with the computer since it's probably maxed out taking in the video ... Thanks once again for your effort on this

Certainly that's true in most cases, but with the sheer CPU power available in newer systems it's worth a try. It can always be turned off if there is a problem with it.

You're welcome! I hope to get some time to make further progress soon.

Igor Babic · January 3rd, 2007, 03:43 AM

Rob, I am very glad that you are considering sound in your app and I agree with you that new motherboards have enough power to do that.
Here is some links to a system components that I want to put together:

http://www.spectra.de/produktliste.c...dummy=1#114781

http://www.spectra.de/produktliste.c...dummy=1#113878

http://www.seagate.com/docs/pdf/data...mentus7200.pdf

http://www.presonus.com/firebox.html

Rob, John can you put some footage online where we can see rolling shutter artefacts? How much motion blur disapear when you max settings against rolling shutter artefacts?

Cesar, have you receive footage from AVT Marlin F131C? That little FW thing also look promising and have global shutter with 2/3"cmos, and it is in our "price range"...

Cesar Rubio · January 3rd, 2007, 09:09 AM

Quote:

Originally Posted by Igor Babic

Cesar, have you receive footage from AVT Marlin F131C? That little FW thing also look promising and have global shutter with 2/3"cmos, and it is in our "price range"...

Igor:

I am still waiting, like everybody else. The Marlin F-131 has a 1/2 CMOS sensor, what do you mean by "FW"?

Thanks,
Cesar Rubio.

Cesar Rubio · January 3rd, 2007, 09:32 AM

Probably you meant FireWire...

Igor Babic · January 3rd, 2007, 11:17 AM

Cesar, AVT say its 2/3" cmos global shutter, 1280x1024, 6,7 um x 6,7 um cell size, raw8 at full res 25fps only...

http://www.alliedvisiontec.com/produ...=18&a=selectid

Cesar Rubio · January 3rd, 2007, 12:04 PM

My bad Igor, you are right...I actually have one F-131c at home!..The F-131 is only capable of shooting 25 fps at FULL resolution, but if you go 1280x720 or 720p in other word, you are able to reach 30p easily.The F-131 has a 10 bit interface.. I will get to you about that later.

Thanks,
Cesar Rubio.

Jason Rodriguez · January 3rd, 2007, 10:56 PM

BTW, Just curious, while the Sumix and assoicated Micron sensor is a good camera, has anyone taken a look at the Kodak CCD cameras?

For instance, the Pulnix or the Imperx cameras using the KAI-4021 4Mpix CCD?

I'm not sure on the pricing, but they are native gig-e cameras using the Pleora gige-link libraries.

Because they are IT CCD's, they are global shuttering, so no problems with rolling shutter like you get on the CMOS chips . . . it's also a fairly large sensor (16mm x 16mm).

Additionally the Imperx camera has the ability to-do in-camera flat field correction and black-frame correction.

Just some thoughts . . . I'm sure they're quite a bit more expensive than the Sumix, but they do look very good. You can see some image samples from the KAI-4021 here:

http://www.illunis.com/images/eric1_35.jpg

http://www.illunis.com/images/Dave%20at%20work.jpg

While the Illunis cameras look interesting, there seems to be some issues with tap matching that I can see in their images, which will give you the infamous "split-screen" effect that we've see with the JVC camera (early models). Imperx with their on-camera flat field correction supposedly solves the issues with any split-screen (tap-mismatching) effects. I guess if someone was interested, they could talk to Illunis and see what the issues might be with tap-mistmatch effects. One thing to keep in mind though is that the Illunis is not a gigabit ethernet camera.

Here's the link to the Imperx:

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

Allied has their Pike F-421B camera that also uses the Kodak CCD and uses firewire 800:

http://www.alliedvisiontec.com/produ...=78&a=selectid

One warning with any CCD camera system that is multi-tap, especially in the high-speed arena, you want to make sure they they adaquately handle tap mismatching in regards to gain offsets (so offsets that occur because of differing signal gains on the two ADC's when signal is added, rather than offsets that are basic DC-level offsets which can be subtracted out), or else you're going to end up with a split-screen effect.

In CMOS cameras, ACD mismatching effects and gain offsets present themselves as column-to-column variations . . . on good CMOS chips, only a subtraction is needed since there's only a pure DC offset between the ADC's . . . on other CMOS chips a second gain correction (flat fielding) is necessary since the ADC's exhibit a gain offset as you add signal. Fill Factory's CMOS chips are notorious for requiring a two-point correction. Micron on the other-hand has a very good process and on-chip correction, so they're actually able to handle these issues on the chip itself, or at the very least an easy black subtraction is all that's necessary to correct any column-to-column offset issues.

Hope this helps,

Jason

Igor Babic · January 4th, 2007, 06:18 AM

Quote:

Originally Posted by Cesar Rubio

My bad Igor, you are right...I actually have one F-131c at home!..The F-131 is only capable of shooting 25 fps at FULL resolution, but if you go 1280x720 or 720p in other word, you are able to reach 30p easily.The F-131 has a 10 bit interface.. I will get to you about that later.

Thanks,
Cesar Rubio.

Thats what I ment all the way (1280x720 raw8/10 25/30p)...Id like to see some footage. Can you do something like 720x576 50/60p?

Jason, Pike F-421B is about EUR 4990. Way to heavy, but good to know stuff about CCD/CMOS.

Jason Rodriguez · January 4th, 2007, 02:22 PM

BTW, While the color reproduction is not too great (and they're a bit noisy as well), Fill Factory's (now Cypress) IBIS-5A sensors do provide a global shutter to prevent any rolling shutter artifacts, and you can get cameras based on that sensor for a bit cheaper as well . . . in fact I believe that Sumix makes a camera using the IBIS-5A, but again, it's downside is that it's not a very pretty chip (not as pretty as the Micron). Also like other Fill Factory sensors, I think it needs a 2-point correction (so both gain and offset) in order to look it's best.

December 30th, 2006, 03:38 PM	#31
Cesar Rubio Regular Crew Join Date: May 2006 Location: Guadalajara, Mexico Posts: 48	Igor: With Stream Pix you DON'T loose any video quality recording in RAW mode, on the contrary you can SAVE all the information that the sensor CCD or CMOS is able to offer, and you can get the BEST video quality in post later on. With SP you also can record audio in sync with your video...give it a try, it's good. Cesar Rubio. __________________ It's the long term experience of problem solving what makes you successful! & what we do in life, echoes in eternity.

January 3rd, 2007, 09:32 AM	#40
Cesar Rubio Regular Crew Join Date: May 2006 Location: Guadalajara, Mexico Posts: 48	Probably you meant FireWire... __________________ It's the long term experience of problem solving what makes you successful! & what we do in life, echoes in eternity.

January 3rd, 2007, 12:04 PM	#42
Cesar Rubio Regular Crew Join Date: May 2006 Location: Guadalajara, Mexico Posts: 48	My bad Igor, you are right...I actually have one F-131c at home!..The F-131 is only capable of shooting 25 fps at FULL resolution, but if you go 1280x720 or 720p in other word, you are able to reach 30p easily.The F-131 has a 10 bit interface.. I will get to you about that later. Thanks, Cesar Rubio. __________________ It's the long term experience of problem solving what makes you successful! & what we do in life, echoes in eternity.

January 1st, 2007, 11:37 AM	#32
John Wyatt Regular Crew Join Date: Jan 2006 Location: West Country, UK Posts: 141	Igor -- the M73 (like many other box cameras) does not record sound in any way, consequently there is no requirement for the capture software to record sound. My own plan for sound is to record it with a DV camcorder (16-bit stereo) running at the same time -- marry up sound with low res proxy later for off-line editing. This sort of set-up (and the need for a sound synching method) may not be suitable for your project if you need high levels of mobility and informal shooting with long takes. Cesar -- it's great to see someone championing 3D! You mentioned StreamPix capture software; I looked at this about two years ago but was put off by the cost. It seemed too expensive for no budget projects like mine; has it come down in price lately? Rob -- can you tell me if you have been able to capture 10-bit images from the M73 with your software? That is an area which really interests me, because it takes the camera even further away from an HDV camcorder. Happy New Year, everyone.

January 2nd, 2007, 08:40 PM	#36
John Wyatt Regular Crew Join Date: Jan 2006 Location: West Country, UK Posts: 141	Rob -- thanks for the latest info: "To reduce artifacts enough with the M73 you'd need to stick with 8 bits" -- shame, 10-bit would have been the cherry on top; I should be more grateful though for what we do have with 8-bit! Rob, do you have any recording durations yet; I was wondering how much longer shooting duration might be compared to the supplied software? Also interested in your opinion of the deBayer you're using compared to the Laplacian one with the supplied software. "I am going to try capturing simultaneous audio in ObscuraCap" -- sounds good Rob; perceived wisdom is it's best not to do anything else with the computer since it's probably maxed out taking in the video (and you don't want any background interrupts to that), though if it can be done (instead of using a separate device), that's obviously a good thing for portability. Thanks once again for your effort on this...

January 3rd, 2007, 03:43 AM	#38
Igor Babic Regular Crew Join Date: Dec 2006 Location: Zagreb, Croatia Posts: 171	Rob, I am very glad that you are considering sound in your app and I agree with you that new motherboards have enough power to do that. Here is some links to a system components that I want to put together: http://www.spectra.de/produktliste.c...dummy=1#114781 http://www.spectra.de/produktliste.c...dummy=1#113878 http://www.seagate.com/docs/pdf/data...mentus7200.pdf http://www.presonus.com/firebox.html Rob, John can you put some footage online where we can see rolling shutter artefacts? How much motion blur disapear when you max settings against rolling shutter artefacts? Cesar, have you receive footage from AVT Marlin F131C? That little FW thing also look promising and have global shutter with 2/3"cmos, and it is in our "price range"...

January 3rd, 2007, 11:17 AM	#41
Igor Babic Regular Crew Join Date: Dec 2006 Location: Zagreb, Croatia Posts: 171	Cesar, AVT say its 2/3" cmos global shutter, 1280x1024, 6,7 um x 6,7 um cell size, raw8 at full res 25fps only... http://www.alliedvisiontec.com/produ...=18&a=selectid

January 3rd, 2007, 10:56 PM	#43
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	BTW, Just curious, while the Sumix and assoicated Micron sensor is a good camera, has anyone taken a look at the Kodak CCD cameras? For instance, the Pulnix or the Imperx cameras using the KAI-4021 4Mpix CCD? I'm not sure on the pricing, but they are native gig-e cameras using the Pleora gige-link libraries. Because they are IT CCD's, they are global shuttering, so no problems with rolling shutter like you get on the CMOS chips . . . it's also a fairly large sensor (16mm x 16mm). Additionally the Imperx camera has the ability to-do in-camera flat field correction and black-frame correction. Just some thoughts . . . I'm sure they're quite a bit more expensive than the Sumix, but they do look very good. You can see some image samples from the KAI-4021 here: http://www.illunis.com/images/eric1_35.jpg http://www.illunis.com/images/Dave%20at%20work.jpg While the Illunis cameras look interesting, there seems to be some issues with tap matching that I can see in their images, which will give you the infamous "split-screen" effect that we've see with the JVC camera (early models). Imperx with their on-camera flat field correction supposedly solves the issues with any split-screen (tap-mismatching) effects. I guess if someone was interested, they could talk to Illunis and see what the issues might be with tap-mistmatch effects. One thing to keep in mind though is that the Illunis is not a gigabit ethernet camera. Here's the link to the Imperx: http://www.imperx.com/machine_vision...meg/index.html Allied has their Pike F-421B camera that also uses the Kodak CCD and uses firewire 800: http://www.alliedvisiontec.com/produ...=78&a=selectid One warning with any CCD camera system that is multi-tap, especially in the high-speed arena, you want to make sure they they adaquately handle tap mismatching in regards to gain offsets (so offsets that occur because of differing signal gains on the two ADC's when signal is added, rather than offsets that are basic DC-level offsets which can be subtracted out), or else you're going to end up with a split-screen effect. In CMOS cameras, ACD mismatching effects and gain offsets present themselves as column-to-column variations . . . on good CMOS chips, only a subtraction is needed since there's only a pure DC offset between the ADC's . . . on other CMOS chips a second gain correction (flat fielding) is necessary since the ADC's exhibit a gain offset as you add signal. Fill Factory's CMOS chips are notorious for requiring a two-point correction. Micron on the other-hand has a very good process and on-chip correction, so they're actually able to handle these issues on the chip itself, or at the very least an easy black subtraction is all that's necessary to correct any column-to-column offset issues. Hope this helps, Jason

January 4th, 2007, 02:22 PM	#45
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	BTW, While the color reproduction is not too great (and they're a bit noisy as well), Fill Factory's (now Cypress) IBIS-5A sensors do provide a global shutter to prevent any rolling shutter artifacts, and you can get cameras based on that sensor for a bit cheaper as well . . . in fact I believe that Sumix makes a camera using the IBIS-5A, but again, it's downside is that it's not a very pretty chip (not as pretty as the Micron). Also like other Fill Factory sensors, I think it needs a 2-point correction (so both gain and offset) in order to look it's best.