New DIY HD Cinema Camera Project

Take Vos · July 18th, 2007, 11:17 AM

John,

I read a couple of months ago that good debayer algorithms can restore better than 90% of the full resolution.

10:9:9 we should call it.

Jose A. Garcia · July 18th, 2007, 12:38 PM

So what's the most acurate debayer algorithm?

Jamie Varney · July 18th, 2007, 01:13 PM

It seems in the open source world the two biggest algorithms are AHD and VNG, both are used by the program Dcraw. It would seem that AHD is better at doing hard edges but is slightly noisy in the large solid areas. VNG is less noisy but has a tendency to produce a 'zipper effect' on edges. This page explains it pretty well: http://www.aim-dtp.net/aim/digicam/d...comparison.htm

Since Dcraw is open source it would be fairly easy to make an AVIsynth plug-in based on one of these two algorithms. A clever coder may even be able to write an AVS script that uses both AHD and VNG together to get the best of both worlds... but it would be extremely slow.

Take Vos · July 18th, 2007, 03:50 PM

Jamie,

I am running AHD at a pretty respectable speed, at around 10 fps 1080x750 with 16 bit half precision floats.

I am not sure how you would combine the two algorithms. Maybe use VNG as a third alternative to horizontal and vertical interpolation. Then make a homogeneity map for all three and select the best one.

There is a paper out there that compares 8 or 10 different algorithms in quality. Interestingly enough although there was a statistically better algorithm than AHD, AHD still visually had less artifacts.

Steven Mingam · July 18th, 2007, 04:20 PM

Quote:

Originally Posted by Jamie Varney

Since Dcraw is open source it would be fairly easy to make an AVIsynth plug-in based on one of these two algorithms. A clever coder may even be able to write an AVS script that uses both AHD and VNG together to get the best of both worlds... but it would be extremely slow.

No because Dcraw source code is an utterly piece of crap that only alien can understand. I know, i was exactly thinking the same... It was faster to read the AHD paper and translate the math to C code. In fact i switched to an other algorithm called "Directional Filtering with A Posteriori Decision" for various reasons already explained in this thread. The avisynth filter currently run at 13 fps @ 1920x800 on my 1.8ghz c2d but i have plenty of room for improvement, let's the MMX magic take over tomorow if i have time ;)
(btw "my" algorithm looked better than AHD in the paper describing it)

Jamie Varney · July 18th, 2007, 04:34 PM

You have any plans to release your filter Steven? I would love to have a look at the results.

Agustin Vrljicak · July 18th, 2007, 09:21 PM

Hi, I've been reading the thread day by day and I'm really excited with what you guys have achieved. I hope you keep the progress until the camera is fully functional. As I have nothing to help you with ('cause I know nothing about algorithms and coding), I just wanted to ask: What size of a sensor would it be needed, to use a 35 mm lens, with no adaptor? I've seen some sensors bigger than the one you Jose are using (1/2" and 1"). Of course I am not suggesting you change board, but it might be useful to think about that. Maybe in the future, a camera with no adapter, will be possible with a larger sensor.
Keep on the work! What you are doing is amazing!!
Agustin

Jose A. Garcia · July 19th, 2007, 04:37 AM

Hi Agustin,

If you read through the thread a bit more you'll find there's a point where we discussed about having a 2/3" Ibis sensor to get a shallower DOF but sacrifying resolution (cause we could only get 720p) or going for full 2K but using a smaller sensor and an adaptor. Now we're using a 1/2.5" cmos and I'll build an adaptor as soon as I have some time.

Now, what's the perfect size to get 35mm DOF? Well, I don't have it clear yet. Some say the sensor must be 36mmx24mm so it has the exact same size of a 35mm negative, but others say 35mm movie negatives are a bit smaller and closer to 1".

But then you have the SI-2K which has a 2/3" sensor and gives some very good DOF without using any adaptor. That's why I started thinking about the Ibis. But anyway I found I really like the Micron sensor. You get a very good looking image with it and you can always build an adaptor to have your own digital cinema 2k camera for less than $2000.

What's the problem using larger cmos? Money and framerate. Most 1" cmos are too expensive or can't get to 24fps. If it was easy or cheap to have a 1" 2k sensor demo board going at 24fps, we'd be using it right now. Altasens gets to 2/3" 1080p but a single sensor costs the same as the whole board I'm using. Also they don't support small projects. I mailed them like four times when I started this. I even told them I was going to start selling the cameras if everything went ok, but all I got was this guy almost telling me not to mail them again.

So it's not as easy as it seems. Anyway I said it in my second or third post here. This project is alive. If I can buy better parts in time, I'll do it.

Wayne Morellini · July 19th, 2007, 07:40 AM

Size apparently doesn't really affect price that much, except marketers would like you to believe that. You can get bigger sensors cheaper than other smaller ones, it is investment and production runs that have an great effect, and the marketing of technology/features. The cost of the extra size can land up being only an small fraction of price, if these things are happening.

Wayne Morellini · July 19th, 2007, 07:48 AM

Quote:

Originally Posted by Sergey Zagorskiy

Hi John,
I'm Petrovich from Sumix Corp. My message relates to movie creation and the problem with long smooth recording of video stream.
Sumix is going to release new soft allowing lossless record of bayer (raw) frames to HD at 1280x1024 with 26fps, 1600x1200 with 19 fps
Recommended to use a separate unfragmeted HDD partition.
The recording is limited by available HD space.
The recorded file is supposed to be processed with special utility, which plays and converts it into avi. The format of the recorded file is opened so one can use his own ideas on compressing etc

in case of questions: zsp@sumix.com

Best regards
Petrovich
Sumix Support team

Long time no hear Sergey. What ever happened with the cinema cameras you were developing for the first group around here?

It is good to see that you have finally published the smooth motion software recording. is it available for the Ibis cameras, and have they improved in S/N ratio sensitivity and performance, or are they exactly the same as 2 years ago?

Thanks

Wayne.

Wayne Morellini · July 19th, 2007, 07:49 AM

Quote:

Originally Posted by Steven Mingam

Btw, thanks to Juan to point me to some pretty interesting and cheap hardware dev kit, i think you can build a usb/ethernet camera head for something like 500$ (or less, depending on the price of the sensor board). Some software developpement is needed but nothing insurmountable...

Steve, what was this hardware?

Wayne Morellini · July 19th, 2007, 07:57 AM

Some cheap hardware:

http://www.brighthand.com/default.asp?newsID=13177

The future of professional production:

http://www.digitalcamcordernews.com/...nsor-developed

8Mp, 2Mp, 1Mp all bin into sensors like these. If sensor technology is good enough, than this resolution might be choice for high end professional use.

Jose A. Garcia · July 19th, 2007, 08:50 AM

Just one thing... I don't think that mini laptop can handle 2k res at 24fps. Or I'm wrong? Because for $250 and an external usb disk to store the clips, it'd be perfect!

Jose A. Garcia · July 19th, 2007, 09:04 AM

What do you think about this:

http://minipc.aopen.com/Global/spec.htm

Cole McDonald · July 19th, 2007, 11:30 AM

Quote:

Originally Posted by Jose A. Garcia

Just one thing... I don't think that mini laptop can handle 2k res at 24fps. Or I'm wrong? Because for $250 and an external usb disk to store the clips, it'd be perfect!

For video, you'll want Firewire, USB2.0 is better on burst transmissions, FW is better on sustained read and write. FW also lets the devices negotiate their own priority on the bus, with USB, the slowest link in the chain is the speed of the rest of it.

July 18th, 2007, 11:17 AM	#181
Take Vos Major Player Join Date: Mar 2007 Location: Amsterdam The Netherlands Posts: 200	John, I read a couple of months ago that good debayer algorithms can restore better than 90% of the full resolution. 10:9:9 we should call it. __________________ VOSGAMES, http://www.vosgames.nl/ developer of Boom Recorder and Mirage Recorder

July 18th, 2007, 03:50 PM	#184
Take Vos Major Player Join Date: Mar 2007 Location: Amsterdam The Netherlands Posts: 200	Jamie, I am running AHD at a pretty respectable speed, at around 10 fps 1080x750 with 16 bit half precision floats. I am not sure how you would combine the two algorithms. Maybe use VNG as a third alternative to horizontal and vertical interpolation. Then make a homogeneity map for all three and select the best one. There is a paper out there that compares 8 or 10 different algorithms in quality. Interestingly enough although there was a statistically better algorithm than AHD, AHD still visually had less artifacts. __________________ VOSGAMES, http://www.vosgames.nl/ developer of Boom Recorder and Mirage Recorder

July 18th, 2007, 12:38 PM	#182
Jose A. Garcia Major Player Join Date: May 2007 Location: Sevilla (Spain) Posts: 439	So what's the most acurate debayer algorithm?

July 18th, 2007, 01:13 PM	#183
Jamie Varney Regular Crew Join Date: Jun 2006 Location: Columbus Ohio Posts: 36	It seems in the open source world the two biggest algorithms are AHD and VNG, both are used by the program Dcraw. It would seem that AHD is better at doing hard edges but is slightly noisy in the large solid areas. VNG is less noisy but has a tendency to produce a 'zipper effect' on edges. This page explains it pretty well: http://www.aim-dtp.net/aim/digicam/d...comparison.htm Since Dcraw is open source it would be fairly easy to make an AVIsynth plug-in based on one of these two algorithms. A clever coder may even be able to write an AVS script that uses both AHD and VNG together to get the best of both worlds... but it would be extremely slow.

July 18th, 2007, 04:34 PM	#186
Jamie Varney Regular Crew Join Date: Jun 2006 Location: Columbus Ohio Posts: 36	You have any plans to release your filter Steven? I would love to have a look at the results.

July 18th, 2007, 09:21 PM	#187
Agustin Vrljicak New Boot Join Date: Jun 2007 Location: Orinda, CA Posts: 15	Hi, I've been reading the thread day by day and I'm really excited with what you guys have achieved. I hope you keep the progress until the camera is fully functional. As I have nothing to help you with ('cause I know nothing about algorithms and coding), I just wanted to ask: What size of a sensor would it be needed, to use a 35 mm lens, with no adaptor? I've seen some sensors bigger than the one you Jose are using (1/2" and 1"). Of course I am not suggesting you change board, but it might be useful to think about that. Maybe in the future, a camera with no adapter, will be possible with a larger sensor. Keep on the work! What you are doing is amazing!! Agustin

July 19th, 2007, 04:37 AM	#188
Jose A. Garcia Major Player Join Date: May 2007 Location: Sevilla (Spain) Posts: 439	Hi Agustin, If you read through the thread a bit more you'll find there's a point where we discussed about having a 2/3" Ibis sensor to get a shallower DOF but sacrifying resolution (cause we could only get 720p) or going for full 2K but using a smaller sensor and an adaptor. Now we're using a 1/2.5" cmos and I'll build an adaptor as soon as I have some time. Now, what's the perfect size to get 35mm DOF? Well, I don't have it clear yet. Some say the sensor must be 36mmx24mm so it has the exact same size of a 35mm negative, but others say 35mm movie negatives are a bit smaller and closer to 1". But then you have the SI-2K which has a 2/3" sensor and gives some very good DOF without using any adaptor. That's why I started thinking about the Ibis. But anyway I found I really like the Micron sensor. You get a very good looking image with it and you can always build an adaptor to have your own digital cinema 2k camera for less than $2000. What's the problem using larger cmos? Money and framerate. Most 1" cmos are too expensive or can't get to 24fps. If it was easy or cheap to have a 1" 2k sensor demo board going at 24fps, we'd be using it right now. Altasens gets to 2/3" 1080p but a single sensor costs the same as the whole board I'm using. Also they don't support small projects. I mailed them like four times when I started this. I even told them I was going to start selling the cameras if everything went ok, but all I got was this guy almost telling me not to mail them again. So it's not as easy as it seems. Anyway I said it in my second or third post here. This project is alive. If I can buy better parts in time, I'll do it.

July 19th, 2007, 07:40 AM	#189
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Size apparently doesn't really affect price that much, except marketers would like you to believe that. You can get bigger sensors cheaper than other smaller ones, it is investment and production runs that have an great effect, and the marketing of technology/features. The cost of the extra size can land up being only an small fraction of price, if these things are happening.

July 19th, 2007, 07:57 AM	#192
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Some cheap hardware: http://www.brighthand.com/default.asp?newsID=13177 The future of professional production: http://www.digitalcamcordernews.com/...nsor-developed 8Mp, 2Mp, 1Mp all bin into sensors like these. If sensor technology is good enough, than this resolution might be choice for high end professional use.

July 19th, 2007, 08:50 AM	#193
Jose A. Garcia Major Player Join Date: May 2007 Location: Sevilla (Spain) Posts: 439	Just one thing... I don't think that mini laptop can handle 2k res at 24fps. Or I'm wrong? Because for $250 and an external usb disk to store the clips, it'd be perfect!

July 19th, 2007, 09:04 AM	#194
Jose A. Garcia Major Player Join Date: May 2007 Location: Sevilla (Spain) Posts: 439	What do you think about this: http://minipc.aopen.com/Global/spec.htm