Ad
Ad
Ad
Pages: [1] 2 3 ... 10 »   Bottom of Page
Print
Author Topic: This is why no RAW on the G7  (Read 69791 times)
Tim Gray
Sr. Member
****
Offline Offline

Posts: 2002



WWW
« on: November 30, 2006, 10:10:21 AM »
ReplyReply

http://news.com.com/2061-10801_3-6139374.h...3-0-5&subj=news

Quotes from both Michael and Chuck Westfall...

Make any sense?
Logged
Paul Sumi
Sr. Member
****
Offline Offline

Posts: 1217


« Reply #1 on: November 30, 2006, 10:52:13 AM »
ReplyReply

Quote
http://news.com.com/2061-10801_3-6139374.h...3-0-5&subj=news

Quotes from both Michael and Chuck Westfall...

Make any sense?
[a href=\"index.php?act=findpost&pid=87867\"][{POST_SNAPBACK}][/a]

Well, "noise" is only one aspect of image quality.  RAW certainly has other advantages which do not need to be rehashed.  The underlying assumption of  Chuck Westfall's comments seem to be that users of this camera will not want to do any post processing.  

Paul
Logged

xmishx
Newbie
*
Offline Offline

Posts: 7


« Reply #2 on: November 30, 2006, 11:11:20 AM »
ReplyReply

Noise on the G series cameras has always been an issue!
Logged

Nill Toulme
Sr. Member
****
Offline Offline

Posts: 741



WWW
« Reply #3 on: November 30, 2006, 11:35:05 AM »
ReplyReply

"...the company now has offered an explanation for the move: increasing the number of megapixels led to more noise per pixel and meant raw was no better than JPEG."

Oh good grief...

Chuck would never have said such a thing.  It must have been his evil twin from Marketing.

Nill
~~
www.toulme.net
« Last Edit: November 30, 2006, 11:39:09 AM by Nill Toulme » Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #4 on: November 30, 2006, 11:57:33 AM »
ReplyReply

Quote
Chuck would never have said such a thing.  It must have been his evil twin from Marketing.

Or his mind is being controlled by aliens...

It smells like a bunch of BS to me; there's no way that the in-camera computer can do a better job auto-processing RAWs than a desktop under the control of a knowledgeable, experienced human. They must think we're a bunch of retards.
Logged

John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #5 on: November 30, 2006, 06:13:13 PM »
ReplyReply

Quote
http://news.com.com/2061-10801_3-6139374.h...3-0-5&subj=news

Quotes from both Michael and Chuck Westfall...

Make any sense?
[a href=\"index.php?act=findpost&pid=87867\"][{POST_SNAPBACK}][/a]

None at all.

Pure nonsense.  Raw will always offer better options for conversion and taming noise.
Logged
dkusner
Newbie
*
Offline Offline

Posts: 14


WWW
« Reply #6 on: December 01, 2006, 11:55:34 PM »
ReplyReply

Quote
None at all.

Pure nonsense.  Raw will always offer better options for conversion and taming noise.
[a href=\"index.php?act=findpost&pid=87956\"][{POST_SNAPBACK}][/a]

It would be sad if this really is Canon's rationale. RAW has never been specifically about image quality for me. It's simply about wanting to use a better RAW converter than the camera, and being able to perform the conversion from the original data as often as I liked rather than having one shot at getting it right. I don't really care how good or bad the JPG image is, or whether there is any discernible difference between a perfect JPG straight from the camera and a good RAW conversion. It's about using the best tool for the job, plus repeatibility.

So the G7's lack of RAW isn't strictly a profitability or product placement decision. They're just embarassed by what happens when you cram 10 MP into the size of a fingernail and can't understand that there's more to RAW than pulling out shadow detail or handling a little overexposure.

I also suspect that Canon is just way too preoccupied with sensors and image quality, rather than on cameras as photographic tools. This is why Nikon has been able to regain so much ground in low- to midrange DSLRs. I don't fault Canon's selection of priorities, but when Nikon catches up in image quality, Canon better have caught up in features and ergonomics.
Logged
Paulo Bizarro
Sr. Member
****
Offline Offline

Posts: 1517


WWW
« Reply #7 on: December 02, 2006, 12:59:57 AM »
ReplyReply

Then give me a G7 with 7 or 8 megapixels AND RAW!
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #8 on: December 02, 2006, 06:52:55 AM »
ReplyReply

Quote
http://news.com.com/2061-10801_3-6139374.h...3-0-5&subj=news

Quotes from both Michael and Chuck Westfall...

Make any sense?
[{POST_SNAPBACK}][/a]

It makes some sense. The smaller pixel has less dynamic range and requires fewer bits to encode the useful information captured by the sensor. It could be that 8 bits at a gamma of 2.2 is sufficient to record what the sensor is capable of capturing.

However, sensors with a higher dynamic range will require more bits to encode their dynamic range, and high end cameras may need 16 bit AD converters. Look at Figure 4 on Roger Clark's web site:

[a href=\"http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/]http://www.clarkvision.com/imagedetail/dig...rmance.summary/[/url]

The other advantages of RAW, such as the ability to use a better decoder on a more powerful computer have been mentioned. One can adjust color temperature after the fact with Lightroom, but having only 8 bits to work with can be a limitation.
Logged
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #9 on: December 02, 2006, 07:51:55 AM »
ReplyReply

Quote
It makes some sense. The smaller pixel has less dynamic range and requires fewer bits to encode the useful information captured by the sensor. It could be that 8 bits at a gamma of 2.2 is sufficient to record what the sensor is capable of capturing.

8 bits with a gamma of 2.2 can record more DR at the pixel level than 12-bit linear, by about 5.5 stops.  In fact, the Leica M8 uses this format for its DNG files.

The real issues are the JPEG artifacts, the compression of shadows and highlights, and the clipping of highlights.

Quote
However, sensors with a higher dynamic range will require more bits to encode their dynamic range, and high end cameras may need 16 bit AD converters. Look at Figure 4 on Roger Clark's web site:

I think Roger's point was that the sensors themselves in recent Canons are capable of warranting 16-bit readout at ISO 100, but just changing the ADC to a 16-bit one is not enough, as there is a lot of read noise at ISO 100.  The conclusion is based on the lower (electron-unit) noise at ISO 1600.  Roger's experiment was a response to a post of mine on usenet where I reported the absolute noise at ISO 1600 was lower than ISO 100.  At that time, I thought that the deficiency of ISO 100 was the bit depth, and I argued that higher bit-depth was the solution, but I realized afterward that posterization doesn't cause that kind of noise; in fact, any noise that shows up as a standard deviation is not caused by posterization.  It became quite clear that bit depth was not the real DR limitation at ISO 100 when I took the same shot with the same manual settings at ISOs 100 and 1600 on my 20D, and then posterized the ISO 1600 image to the same number of levels as the ISO 100 image, and it gained only a small amount of visible noise, and was still orders of magnitude clearer than the ISO 100 image.

A system with only 10 bits and no read noise would be better than what we have now, IMO.  However, with no read noise, you might as well digitize at 16 bits, because that will be even better.

I have a lot of experiments in my head that I'd like to try, but never get around to.  One is to take a stack of ISO 100 images (about 16) with my 20D on a tripod, with a 10mm lens for maximum registration.  The result will be a 16-bit linear RAW, with a standard deviation of about 0.52 at the black level, and low in shot noise as well (all noises will be two stops weaker).  Then, compare a single one of the 16 images with the stack, and with the stack posterized to various bit depths.

Quote
The other advantages of RAW, such as the ability to use a better decoder on a more powerful computer have been mentioned. One can adjust color temperature after the fact with Lightroom, but having only 8 bits to work with can be a limitation.
[a href=\"index.php?act=findpost&pid=88225\"][{POST_SNAPBACK}][/a]

I'm not sure exactly how much RAW headroom is clipped in a G7 JPEG, but even if it is half of what the DSLRs clip (typically 1 stop for the DSLRs), you will get less noise by shooting your ISO 400 at ISO 283 with +0.5 EC, something you can not do in JPEG mode without blowing highlights.

Canon scares me, to be honest.  I don't think that people who know what they are doing are making decisions, or they do, but are just trying to save money by not having to answer tech support calls from people who can't figure out what to do with the RAW shots they took.
Logged
jaj
Newbie
*
Offline Offline

Posts: 4


« Reply #10 on: December 02, 2006, 11:21:12 AM »
ReplyReply

Quote
http://news.com.com/2061-10801_3-6139374.h...3-0-5&subj=news

Quotes from both Michael and Chuck Westfall...

Make any sense?
[a href=\"index.php?act=findpost&pid=87867\"][{POST_SNAPBACK}][/a]

well, the essential question for us buyers of G7 (and recently engaged in learning the control of raw-files) is whether a firm-ware update is possible, or impossible because of particular hardware of the G7?

jaj
Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #11 on: December 02, 2006, 05:27:54 PM »
ReplyReply

Quote
It makes some sense. The smaller pixel has less dynamic range and requires fewer bits to encode the useful information captured by the sensor. It could be that 8 bits at a gamma of 2.2 is sufficient to record what the sensor is capable of capturing.

That's preposterous. Canon is claiming that the processor and firmware inside the camera can do a better job of processing the RAW data from the sensor than you, your much-more-powerful-than-the-camera desktop computer, your favorite RAW converter, and Photoshop. That's just as dumb as claiming that the minilab at the corner drugstore can auto-process and print film images better than an experienced film photographer familiar with developing, processing, and printing film. The fact that small-pixel sensors are noisy means than the experience of the human over the machine is more important, not less. I have gotten some good images from my Olympus SP-350, but I have to put more effort into them than I do images from my DSLRs; the digicam RAWs need more work to meet high technical standards than the DSLR RAWs.
Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #12 on: December 02, 2006, 07:20:41 PM »
ReplyReply

Quote
That's preposterous. Canon is claiming that the processor and firmware inside the camera can do a better job of processing the RAW data from the sensor than you, your much-more-powerful-than-the-camera desktop computer, your favorite RAW converter, and Photoshop.
[a href=\"index.php?act=findpost&pid=88311\"][{POST_SNAPBACK}][/a]

Really, haven't you heard about ASICs (application specific integrates circuits)? They can perform specialized tasks quite rapidly, whereas a general purpose microprocessor such as in a desktop uses less optimized code for demosiacing and other graphics functions. This point was made by Andrew Rodney in response to a previous post that I had made voicing the same argument as yours above. Do you have any data to support your assertions?
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #13 on: December 02, 2006, 07:48:24 PM »
ReplyReply

Quote
8 bits with a gamma of 2.2 can record more DR at the pixel level than 12-bit linear, by about 5.5 stops.  In fact, the Leica M8 uses this format for its DNG files.

I think Roger's point was that the sensors themselves in recent Canons are capable of warranting 16-bit readout at ISO 100, but just changing the ADC to a 16-bit one is not enough, as there is a lot of read noise at ISO 100.  The conclusion is based on the lower (electron-unit) noise at ISO 1600.  Roger's experiment was a response to a post of mine on usenet where I reported the absolute noise at ISO 1600 was lower than ISO 100.  At that time, I thought that the deficiency of ISO 100 was the bit depth, and I argued that higher bit-depth was the solution, but I realized afterward that posterization doesn't cause that kind of noise; in fact, any noise that shows up as a standard deviation is not caused by posterization.  It became quite clear that bit depth was not the real DR limitation at ISO 100 when I took the same shot with the same manual settings at ISOs 100 and 1600 on my 20D, and then posterized the ISO 1600 image to the same number of levels as the ISO 100 image, and it gained only a small amount of visible noise, and was still orders of magnitude clearer than the ISO 100 image.

A system with only 10 bits and no read noise would be better than what we have now, IMO.  However, with no read noise, you might as well digitize at 16 bits, because that will be even better.

[{POST_SNAPBACK}][/a]

Yes, posterization is caused when there are not enough bits to quantize the number of levels in the image and the gaps in the data are too wide and the differences are obvious to the eye. Random noise will tend to hide posterization

I don't understand your statement that ISO 1600 has less noise than ISO 100. This defies common experience. The read noise is higher at ISO 100, but as we have gone over before, photon noise is predominant at all levels of a real world photograph and this is much less when you collect a large number of photons at low ISO. I know you don't agree with Roger's analysis, but where are your data?

[a href=\"http://www.clarkvision.com/imagedetail/evaluation-1d2/index.html]http://www.clarkvision.com/imagedetail/eva...-1d2/index.html[/url]

10 bit linear is severely limited in dynamic range as shown by this table on Norman Koren's web site. If you assume the darkest zone should have 8 levels, 10 bit linear has a DR of only 7 stops as shown in Norman's table.

http://www.normankoren.com/digital_tonality.html

Bill
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7776



WWW
« Reply #14 on: December 02, 2006, 08:18:04 PM »
ReplyReply

Quote
That's preposterous. Canon is claiming that the processor and firmware inside the camera can do a better job of processing the RAW data from the sensor than you, your much-more-powerful-than-the-camera desktop computer, your favorite RAW converter, and Photoshop. That's just as dumb as claiming that the minilab at the corner drugstore can auto-process and print film images better than an experienced film photographer familiar with developing, processing, and printing film. The fact that small-pixel sensors are noisy means than the experience of the human over the machine is more important, not less. I have gotten some good images from my Olympus SP-350, but I have to put more effort into them than I do images from my DSLRs; the digicam RAWs need more work to meet high technical standards than the DSLR RAWs.
[a href=\"index.php?act=findpost&pid=88311\"][{POST_SNAPBACK}][/a]

As I wrote in a separate post on the same topic, the Nikon D50 is one example of a camera whose high ISO jpg appears to be very difficult to top even when carefully processing the RAW file in post. Thom Hogan has written on this topic.

Besides, I am not sure that Canon is telling us the real truth here. IMHO, they have just reached the conclusion that most users would not be able to do better with a RAW file than they managed to achieve with a dedicated processor in camera. It doesn't even have to mean that doing better is impossible, just that it is difficult enough that few people would manage, especially with a non Canon RAW converter.

My view is that this is just yet another example of the open/closed system topic. Canon with the G7 is doing the same thing Hasselblad is doing with the H3D. They are deciding to close a system to get more control.

Cheers,
Bernard
Logged

A few images online here!
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #15 on: December 02, 2006, 11:00:51 PM »
ReplyReply

Quote
Yes, posterization is caused when there are not enough bits to quantize the number of levels in the image and the gaps in the data are too wide and the differences are obvious to the eye. Random noise will tend to hide posterization

I don't understand your statement that ISO 1600 has less noise than ISO 100.

I didn't make such a statement.  I said that ISO 1600 has less absolute noise, to distinguish it from relative noise, which is higher at ISO 1600 in all regards.

Quote
This defies common experience. The read noise is higher at ISO 100, but as we have gone over before, photon noise is predominant at all levels of a real world photograph and this is much less when you collect a large number of photons at low ISO.

The photon shot noise is related only to the amount of light falling on the sensor; it has nothing to do with ISO, so when the same light falls on the sensor, ISO 100 and 1600 have the same shot noise, but the readout noise is greater at ISO 100, and hence the total of all absolute noises is higher for 100 (1600 is not worse in any case).


Quote
I know you don't agree with Roger's analysis, but where are your data? 

I don't necessarily disagree with that particular analysis; what I have disagreed with in the past is his conclusion that over-sampling well electrons is not beneficial. I just think that as a scientist, he may not have stressed the point that the non-scientist is likely to miss, that it is the sensor, and not the whle amplification/read-out system, that is capable of fully utilizing 16 bits of precision.

Quote
10 bit linear is severely limited in dynamic range as shown by this table on Norman Koren's web site. If you assume the darkest zone should have 8 levels, 10 bit linear has a DR of only 7 stops as shown in Norman's table.[a href=\"index.php?act=findpost&pid=88328\"][{POST_SNAPBACK}][/a]

Most of these discussions tend to ignore the huge difference between the dynamic range of the pixel as an independent measuring device, and the dynamic range of an image, which is a community of pixels.  You can have far more DR in an image than the bit-depth allows for a single pixel.  How do you think that all of those old dithering schemes worked when we had 256-color and even 16-color displays?

I will show you the results when I get to the experiment, but from what I have generally seen, bit depth is not as big a contributor to image DR as people make it out to be; it's the noise that makes the biggest difference in current systems.  Get noise out of the way, and then bit-depth will be the bigger issue.
Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #16 on: December 03, 2006, 01:38:51 AM »
ReplyReply

Quote
Really, haven't you heard about ASICs (application specific integrates circuits)? They can perform specialized tasks quite rapidly, whereas a general purpose microprocessor such as in a desktop uses less optimized code for demosiacing and other graphics functions. This point was made by Andrew Rodney in response to a previous post that I had made voicing the same argument as yours above. Do you have any data to support your assertions?

Yes, I've heard of ASICs. Yes, they can do certain things faster than CPUs, But CPUs can use more complex algorithms than can be programmed into an ASIC, which means that CPUs can do a better job of accomplishing some tasks than ASICs when the more complex algorithm is more effective. But the ASIC vs. CPU iquestion addresses none of the following:

How many digicams are individually profiled to optimize color accuracy? I get much more accurate color from my SP-350 from ACR after running Tom Fors' Color Checker calibration then I get from the in-camera JPEGs. The same is true of images from my 1Ds and 1D-MkII; calibrated ACR beats the camera JPEGs hands-down.

Since when does any camera, digicam, DSLR, or MFDB, get white balance right 100% of the time? ACR's auto white balance is generally better than the camera's, but still can be fooled by some subject matter. But ACR gives me total freedom to change it, unlike pre-baked JPEGs. I don't care how good the camera is, I know I'm better at picking a white balance than it is. AWB (in-camera and in-converter) gets one in the ballpark, but is far from perfect. There's too many situations where it can be fooled to blindly trust it.

The inherent advantages of high-bit image editing go out the window when shooting JPEGs. 16-bit vs. 8-bit editing has been discussed at length in many threads, so I won't rehash it all here. Why edit with one hand tied behind your back?

Camera JPEGs throw away dynamic range and color gamut; I don't see any camera supporting ProPhoto (and to do so with any 8-bit file format would be foolish anyway, due to the posterization & banding it would cause) even though many sensors can capture colors outside Adobe RGB, including Canon's, and RAW converters are much better at highlight recovery than camera firmware. Camera JPEGs give you <100% of the sensor's DR and gamut, RAW gives you 100% of the sensor gamut and at least 100% of the sensor's DR to work with.

Bottom line: being limited to JPEG constrains and compromises color accuracy, image gamut, dynamic range, white balance, and the ability to apply adjustments to the image without intruducing posterization and banding artifacts, regardless of how good in-camera noise processing may be. And I'm not convinced that camera firmware can do better than Neat Image, Noise Ninja, etc. in that department, either. Do you have any examples of camera JPEGs outperforming these tools?
« Last Edit: December 03, 2006, 01:42:59 AM by Jonathan Wienke » Logged

BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7776



WWW
« Reply #17 on: December 03, 2006, 03:46:04 AM »
ReplyReply

Quote
And I'm not convinced that camera firmware can do better than Neat Image, Noise Ninja, etc. in that department, either. Do you have any examples of camera JPEGs outperforming these tools?
[{POST_SNAPBACK}][/a]

Jonathan,

You might want to check this out:

[a href=\"http://www.bythom.com/D50REVIEW.htm]http://www.bythom.com/D50REVIEW.htm[/url]

Go down to the section called "Imatest Results and Interpretation", and read the 5th bullet on noise.

Especially the last part, that I quote without permission from Thom Hogan (I hope he'll forgive me):

"The bad news is this: the D50, like the D2x, is doing something with JPEGs that doesn't get done with NEFs. NEFs from a D50, while still a better than my D70s at high ISO values, have more noise in them than JPEGs. Detail in JPEGs is a little lower than in NEFs, so perhaps we've got some very good noise reduction in play with the JPEG rendering engine. "

Best regards,
Bernard
Logged

A few images online here!
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #18 on: December 03, 2006, 08:16:20 AM »
ReplyReply

Quote
It smells like a bunch of BS to me; there's no way that the in-camera computer can do a better job auto-processing RAWs than a desktop under the control of a knowledgeable, experienced human. They must think we're a bunch of retards.
[{POST_SNAPBACK}][/a]

I get the distinct impression that the ratio of "retards" vs "knowledgeable, experienced humans" is roughly comparable to "build-in lens" vs "DSLR" in [a href=\"http://www.dpreview.com/news/0612/06120101cipashipq3.asp]this link[/url]...

Some of the DSLR owners are still retards, some of the bil owners are still knowledgeable. Either way, if there is a remote chance that you can end up with a bad name on the retard side, the data suggests that that chance will hurt sales significantly.

Something to ponder: Michael apparently chose the "No RAW, but well build" option over the "6sec RAW, but Fuji build" option. No doubt there are other significant factors, but they may come down to the same conundrum: "No RAW + advantage" over "6sec RAW + disadvantage"...
Logged

Regards,
Oscar Rysdyk
theimagingfactory
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #19 on: December 03, 2006, 08:50:17 AM »
ReplyReply

Quote
Jonathan,

You might want to check this out:

http://www.bythom.com/D50REVIEW.htm

Go down to the section called "Imatest Results and Interpretation", and read the 5th bullet on noise.

Especially the last part, that I quote without permission from Thom Hogan (I hope he'll forgive me):

"The bad news is this: the D50, like the D2x, is doing something with JPEGs that doesn't get done with NEFs. NEFs from a D50, while still a better than my D70s at high ISO values, have more noise in them than JPEGs. Detail in JPEGs is a little lower than in NEFs, so perhaps we've got some very good noise reduction in play with the JPEG rendering engine. "

Best regards,
Bernard

That's because he's not doing any noise reduction whatsoever in his RAW workflow. I saw no mention of Neat Image, Noise Ninja, Noiseware, or any other PS noise reduction plugin. Obviously no noise reduction is done in-camera on RAW files, that's the whole point of RAW--giving the user control of what processing and how much, instead of the camera making all processing decisions. Thom's comparison is between in-camera noise reduction and nothing at all, and if the camera fails that test, the firmware programmers at Nikon ought to be fired en masse.
Logged

Pages: [1] 2 3 ... 10 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad