4:4:4 10bit single CMOS HD project

Obin Olson · July 28th, 2004, 02:58 PM

Rai...look at the image below:

http://obin.weet.us/gamma 2.4.tif

dude...would YOU shoot a feature film with that much artifacting in the darks?

that is a gamma of 2.4...LOTS of dynamic range and LOTS of overexposure artifacts in the blacks..

Rai...I sent you 3 files...they all came back...if you want images then you need to deal with your email account!!!!

don't beg me for stuff just to have it sent back ;)

Steve Nordhauser · July 28th, 2004, 03:02 PM

Rai:
I don't think I understood what you said. There are two separate issues in a rolling shutter camera. The first is exposure which can be set from a maximum of a frame time to a minimum of a couple of rows. This will do as you expect - shorter frame times yield less light in but less motion blur. Second is the rolling shutter artifact - the time difference from the readout of the top of frame to bottom. This is either 1/24th sec or 1/48th sec if you drop every other frame (which limits the max exposure and *may* cause more smearing). I think the max 1/48th sec exposure might be what you referred to.

I'm going to remind everyone that the excitement on the Altasens is still over a rolling shutter sensor - running at 24/30/48/60fps.

Obin Olson · July 28th, 2004, 03:12 PM

painting RAW:

http://obin.weet.us/painting.tif

painting with gamma set at 2.2 and some cc to make it standout:

http://obin.weet.us/painting gamma 2.2 some cc.tif

Rai Orz · July 28th, 2004, 03:24 PM

@Obin, first thanks for the images (email). My people, also in the post, say the RAW images are okay.

---->http://obin.weet.us/gamma 2.4.tif

dude...would YOU shoot a feature film with that much artifacting in the darks?<----

No, but we would also not shoot with a 35mm ARRI this view. Obin, i think there is a differently "look" we need. You compare all images with a normal video camera. But i dont want video look. I love cinema, and so i love your earlier images.
But my main problem now is the rolling shutter

Obin Olson · July 28th, 2004, 03:37 PM

what "look" are you going for ?? High Contrast? low contrast? bright colors?

Rai Orz · July 28th, 2004, 03:47 PM

Steve,
sorry, i mean rolling shutter artifact. You said it: The time difference from readout of the top of frame to bottom is very slow. It lasts a whole frame. And that is exactly the problem. At 24fps its 1/24th sec. At 48fps its 1/48th sec. It is much to slow. A mechanical shutter in a movie camera do this 24times faster.

Obin,
i not the man who make the post.
I can only say what i like. It is more movie "look" than video "look"

video = TV-News, Sports or so.
Movie = The X Files

Obin Olson · July 28th, 2004, 04:04 PM

http://obin.weet.us/outdoors-1.wmv

gamma about 2.0 and minor CC for a natural tone

Jason Rodriguez · July 28th, 2004, 04:19 PM

Quote:

don't think I understood what you said. There are two separate issues in a rolling shutter camera. The first is exposure which can be set from a maximum of a frame time to a minimum of a couple of rows. This will do as you expect - shorter frame times yield less light in but less motion blur. Second is the rolling shutter artifact - the time difference from the readout of the top of frame to bottom. This is either 1/24th sec or 1/48th sec if you drop every other frame (which limits the max exposure and *may* cause more smearing). I think the max 1/48th sec exposure might be what you referred to

Ahh, now I understand!! :-)

So that's how camera manufacturers get around the rolling shutter artifacting problems-the chip is always running at 60Hz (for NTSC), it's just the exposure time is different for the rows, but the chip itself is always reading out at a fast enough "frame-rate" to eliminate any rolling shutter artifacts. And that's why you can't get the shutter slower that 1/60th of a second (in normal mode) on an NTSC camera.

Now the progressive cameras like the DVX100 though can do 1/30th of a second shutters, and they supposedly shoot at 24fps. I haven't seen any rolling shutter artifacts with that camera, and again, I think a lot of the problem has to do with lack of motion blur. When the shutter is all the way open I'm used to seeing much more motion blur than what I'm seeing in these samples. So I'm not sure if the integration time is long enough or not, but I'm figuring that the more motion blur in the frame, the less you're going to see any rolling shutter artifacts.

BTW, how come you can't have the chip read out in 1/48th of a second an then blank for 1/48th of a second? I mean suppose you wanted 32fps or 33fps, how would you do that and still maintain the /148th of a second shutter since they don't evenly divide into each other?

Obin Olson · July 28th, 2004, 04:22 PM

rolling shutter is a thing of CMOS I think..I have never seen an issue with CCD chips..and YES JVC etc will run the Altasense at 60mhz ALL THE TIME and take every other frame to tape for 30fps etc

Ben Syverson · July 28th, 2004, 05:05 PM

The JVC page mentioned it had a max rate of 75mhz -- maybe they're running at that speed all the time.

Rai, what do you need?

Rai Orz · July 28th, 2004, 05:13 PM

Jason and Obin,
it is not only the motion blur. A movie have also motion blur and i like it. But a movie camera have not the rolling shutter artifact problem.
(Oh my bad english...) If you have strings (up to down), or just the manhatten skyline and you move the camera (left to right) the first line in the sensor is readout, than a little time later the next line, and so on, but in the meantime the camera move on. The position is not the same, because it takes 1/24th sec from the first line on the top to the last line.
The picture "swings". I hope you understand

With global shutter, you have also motion blur (and i like this), but you will dont have this artifact swinging problem.

Ben, i think i like your camera. We have some ideas to change the interface and other things, so we can use it (with 10bit and in global shutter mode) next days. Please email...

Ben Syverson · July 28th, 2004, 05:25 PM

Rai, the interface is totally fixed -- it's USB2. And Sumix is working on 10bit software as we speak. What exactly do you want to do?

Steve Nordhauser · July 28th, 2004, 07:00 PM

My boss did a pair of images of the IBIS-5 and the Micron 1.3Mpix using the standard XCAP tools for image enhancement. The IBIS-5 was run in global shutter mode since the color is better that way. It was run at a slower clock to allow for sufficient exposure time.

http://siliconimaging.com/Samples/SI....3%20GG1.2.jpg

http://siliconimaging.com/Samples/SI....3%20GG1.2.jpg

Rai:
You have it. Motion blur and rolling shutter artifacts are totally separate effects other than their visual interaction. The Altasens can be run at any rate up to 60fps.

More on the rolling shutter beast:
The 60fps of NTSC was done with interlacing. First you display a frame of even lines in a 60th of a sec and then a frame of odd. Any rolling shutter effects would be jagged that way. CCDs have been available for years with either interline transfer or full frame transfer - both ways of capturing a full frame at once.

Obin Olson · July 28th, 2004, 07:26 PM

that's why I like the Micron......WAY more color then the IBIS

Ben Syverson · July 28th, 2004, 07:29 PM

Steve,

I Level'd the 1280 image and boosted the chroma by 1.25x -- of course it looks kind of rough because the input was a jpeg, but you can get a sense.

si1280corrected.jpg

I had one version that looked essentially the same as the Micron image, but i figured what the heck -- I'll up the color a little bit more and white balance it a little better. (It's tough because the flourescent is really blue)

- ben

July 28th, 2004, 03:02 PM	#1187
Steve Nordhauser Silicon Imaging, Inc. Join Date: May 2004 Location: Troy, NY USA Posts: 325	Rai: I don't think I understood what you said. There are two separate issues in a rolling shutter camera. The first is exposure which can be set from a maximum of a frame time to a minimum of a couple of rows. This will do as you expect - shorter frame times yield less light in but less motion blur. Second is the rolling shutter artifact - the time difference from the readout of the top of frame to bottom. This is either 1/24th sec or 1/48th sec if you drop every other frame (which limits the max exposure and may cause more smearing). I think the max 1/48th sec exposure might be what you referred to. I'm going to remind everyone that the excitement on the Altasens is still over a rolling shutter sensor - running at 24/30/48/60fps. __________________ Silicon Imaging, Inc. We see the Light! http://www.siliconimaging.com

July 28th, 2004, 07:00 PM	#1198
Steve Nordhauser Silicon Imaging, Inc. Join Date: May 2004 Location: Troy, NY USA Posts: 325	IBIS-5 vs Micron 1.3Mpix My boss did a pair of images of the IBIS-5 and the Micron 1.3Mpix using the standard XCAP tools for image enhancement. The IBIS-5 was run in global shutter mode since the color is better that way. It was run at a slower clock to allow for sufficient exposure time. http://siliconimaging.com/Samples/SI....3%20GG1.2.jpg http://siliconimaging.com/Samples/SI....3%20GG1.2.jpg Rai: You have it. Motion blur and rolling shutter artifacts are totally separate effects other than their visual interaction. The Altasens can be run at any rate up to 60fps. More on the rolling shutter beast: The 60fps of NTSC was done with interlacing. First you display a frame of even lines in a 60th of a sec and then a frame of odd. Any rolling shutter effects would be jagged that way. CCDs have been available for years with either interline transfer or full frame transfer - both ways of capturing a full frame at once. __________________ Silicon Imaging, Inc. We see the Light! http://www.siliconimaging.com

July 28th, 2004, 02:58 PM	#1186
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	Rai...look at the image below: http://obin.weet.us/gamma 2.4.tif dude...would YOU shoot a feature film with that much artifacting in the darks? that is a gamma of 2.4...LOTS of dynamic range and LOTS of overexposure artifacts in the blacks.. Rai...I sent you 3 files...they all came back...if you want images then you need to deal with your email account!!!! don't beg me for stuff just to have it sent back ;)

July 28th, 2004, 03:12 PM	#1188
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	painting RAW: http://obin.weet.us/painting.tif painting with gamma set at 2.2 and some cc to make it standout: http://obin.weet.us/painting gamma 2.2 some cc.tif

July 28th, 2004, 03:24 PM	#1189
Rai Orz Regular Crew Join Date: Jun 2004 Location: Germany Posts: 136	@Obin, first thanks for the images (email). My people, also in the post, say the RAW images are okay. ---->http://obin.weet.us/gamma 2.4.tif dude...would YOU shoot a feature film with that much artifacting in the darks?<---- No, but we would also not shoot with a 35mm ARRI this view. Obin, i think there is a differently "look" we need. You compare all images with a normal video camera. But i dont want video look. I love cinema, and so i love your earlier images. But my main problem now is the rolling shutter

July 28th, 2004, 03:37 PM	#1190
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	what "look" are you going for ?? High Contrast? low contrast? bright colors?

July 28th, 2004, 03:47 PM	#1191
Rai Orz Regular Crew Join Date: Jun 2004 Location: Germany Posts: 136	Steve, sorry, i mean rolling shutter artifact. You said it: The time difference from readout of the top of frame to bottom is very slow. It lasts a whole frame. And that is exactly the problem. At 24fps its 1/24th sec. At 48fps its 1/48th sec. It is much to slow. A mechanical shutter in a movie camera do this 24times faster. Obin, i not the man who make the post. I can only say what i like. It is more movie "look" than video "look" video = TV-News, Sports or so. Movie = The X Files

July 28th, 2004, 04:04 PM	#1192
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	http://obin.weet.us/outdoors-1.wmv gamma about 2.0 and minor CC for a natural tone

July 28th, 2004, 04:22 PM	#1194
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	rolling shutter is a thing of CMOS I think..I have never seen an issue with CCD chips..and YES JVC etc will run the Altasense at 60mhz ALL THE TIME and take every other frame to tape for 30fps etc

July 28th, 2004, 05:05 PM	#1195
Ben Syverson Major Player Join Date: Mar 2004 Location: chicago Posts: 434	The JVC page mentioned it had a max rate of 75mhz -- maybe they're running at that speed all the time. Rai, what do you need?

July 28th, 2004, 05:13 PM	#1196
Rai Orz Regular Crew Join Date: Jun 2004 Location: Germany Posts: 136	Jason and Obin, it is not only the motion blur. A movie have also motion blur and i like it. But a movie camera have not the rolling shutter artifact problem. (Oh my bad english...) If you have strings (up to down), or just the manhatten skyline and you move the camera (left to right) the first line in the sensor is readout, than a little time later the next line, and so on, but in the meantime the camera move on. The position is not the same, because it takes 1/24th sec from the first line on the top to the last line. The picture "swings". I hope you understand With global shutter, you have also motion blur (and i like this), but you will dont have this artifact swinging problem. Ben, i think i like your camera. We have some ideas to change the interface and other things, so we can use it (with 10bit and in global shutter mode) next days. Please email...

July 28th, 2004, 05:25 PM	#1197
Ben Syverson Major Player Join Date: Mar 2004 Location: chicago Posts: 434	Rai, the interface is totally fixed -- it's USB2. And Sumix is working on 10bit software as we speak. What exactly do you want to do?

July 28th, 2004, 07:26 PM	#1199
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	that's why I like the Micron......WAY more color then the IBIS

July 28th, 2004, 07:29 PM	#1200
Ben Syverson Major Player Join Date: Mar 2004 Location: chicago Posts: 434	Steve, I Level'd the 1280 image and boosted the chroma by 1.25x -- of course it looks kind of rough because the input was a jpeg, but you can get a sense. si1280corrected.jpg I had one version that looked essentially the same as the Micron image, but i figured what the heck -- I'll up the color a little bit more and white balance it a little better. (It's tough because the flourescent is really blue) - ben