Ad
Ad
Ad
Pages: « 1 [2] 3 »   Bottom of Page
Print
Author Topic: Canon Highlight Priority  (Read 71967 times)
203
Guest
« Reply #20 on: December 07, 2007, 10:07:12 AM »
ReplyReply

Quote from Chuck Westfall:

Re: Highlight Prio.

Dynamic range of the 40D remains the same as the 30D despite higher resolution and smaller pixels.

What you're seeing with HTP is a modified image processing algorithm combined with the effects of 14-bit A/D conversion of the RAW image data. Reviewers who discuss the benefits of HTP are effectively pointing out how image processing modifications combined with a higher bit-depth in the RAW image data can improve the rendition of highlight details compared to earlier cameras without those features.

And once again, the 14-bit A/D conversion enhances the effect of HTP because it provides 4 times as many gradations to choose from as the RAW image data from cameras with 12-bit A/D conversion.
__________________
Chuck Westfall
Director/Media & Customer Relationship
Camera Division/Canon U.S.A., Inc.
Logged
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #21 on: December 07, 2007, 10:13:25 AM »
ReplyReply

Quote
Quote from Chuck Westfall:

Re: Highlight Prio.
And once again, the 14-bit A/D conversion enhances the effect of HTP because it provides 4 times as many gradations to choose from as the RAW image data from cameras with 12-bit A/D conversion.[a href=\"index.php?act=findpost&pid=158949\"][{POST_SNAPBACK}][/a]

If the 14 bits can't improve DR, it really can't improve anything else, because noise, which is the barrier to DR, is also the barrier to being able to benefit from smooth "gradients".
Logged
203
Guest
« Reply #22 on: December 07, 2007, 10:26:05 AM »
ReplyReply

Quote
If the 14 bits can't improve DR, it really can't improve anything else, because noise, which is the barrier to DR, is also the barrier to being able to benefit from smooth "gradients".
[a href=\"index.php?act=findpost&pid=158953\"][{POST_SNAPBACK}][/a]

It seems that everyone who has reviewed these cameras has stated that the files look better in color, shadow detail, highlight retention, etc.

So, uh, I am going to take Chuck's word over yours :-)
« Last Edit: December 07, 2007, 10:28:19 AM by 203 » Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8845


« Reply #23 on: December 07, 2007, 12:13:55 PM »
ReplyReply

Quote
It seems that everyone who has reviewed these cameras has stated that the files look better in color, shadow detail, highlight retention, etc.

So, uh, I am going to take Chuck's word over yours :-)
[a href=\"index.php?act=findpost&pid=158961\"][{POST_SNAPBACK}][/a]

I think the idea here is that the in-camera processed jpegs look better, not necessarily Raw images that have been underexposed by one stop at a lower ISO and processed accordingly in ACR. Sometimes just one click on 'auto' in ACR can correct an image, whether underexposed or fully exposed to the right.

On the occasions that I've resorted to shooting in jpeg mode, I've found that at least some of the shots have had irrecoverable highlight detail loss, despite my attempts at being careful about exposure.

I think this HTP feature is strictly for jpeg shooters. Of course it works in RAW mode, but I can't think why you would use it in RAW mode.

BTW, an image which is fully exposed to the right in jpeg mode will suffer from irrecoverable loss of highlight detail, without the HTP feature.
« Last Edit: December 07, 2007, 12:19:09 PM by Ray » Logged
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #24 on: December 07, 2007, 12:47:17 PM »
ReplyReply

Quote
It seems that everyone who has reviewed these cameras has stated that the files look better in color, shadow detail, highlight retention, etc.

So, uh, I am going to take Chuck's word over yours :-)
[a href=\"index.php?act=findpost&pid=158961\"][{POST_SNAPBACK}][/a]

Chuck is a business person.  What he is going to say is what he thinks is good for the company (and his own standing there), and within the limits of his own knowledge and understanding.

Canon's official statements have proven to be incorrect in the past; remember the white papers that said how the "big pixel" cameras (5D and 1Dmk2) have better high-ISO performance because they collected more photons with their bigger photosites?  The fact is, these cameras have a lower quantum efficiency than the Canon 20D, and even less than some P&S cameras.  The high-ISO pixel performance is almost exactly the same on all three cameras.  The only reason the 1Dmk2 gives better high-ISO images than the 20D is because the coarser pixel spacing requires less from the optics for a sharp image, and the anti-aliasing filter is weaker, meaning the sharpness of "image" is greater, relative to the sharpness of noise.  Nothing electrical, or related to photons at all.  The bigger pixels captured more photons total, but that requires a greater exposure (ISO 50).

Never trust a company, or its representatives, completely.  Always verify what they say.

As far as 14-bit Canons are concerned, it is impossible to get a 12-bit RAW out of the same camera, so you are comparing 14-bit from one, newer camera, along with the possibility that it is forcing the converter to work with more shadow precision (something it could have always done without the extra two bits), to another, older camera.

Those of us who have taken 14-bit RAW data and quantized the data to 12-bit, have found no loss in image detail in any tonal range.

Do you have any idea how much pixel values vary due to noise alone?  The difference between values at 14-bit, and what they are rounded to when quantized to 12 bits, is only a very tiny fraction of the amount the pixel values are varying due to noise, except in the very deepest shadows of the lowest ISOs.  In fact, for the midtone and highlight areas, the rounding is virtually infinitessimal compared to the shot noise.

All the apparent benefits of 14-bit RAW processing are available by just promoting 12-bit RAWs to 14 bits by padding them with '10' (binary), just like you get better results by converting an 8-bit JPEG to 16-bit in photoshop before extensively editing it.
Logged
203
Guest
« Reply #25 on: December 07, 2007, 01:02:29 PM »
ReplyReply

Quote
As far as 14-bit Canons are concerned, it is impossible to get a 12-bit RAW out of the same camera, so you are comparing 14-bit from one, newer camera, along with the possibility that it is forcing the converter to work with more shadow precision.......
[a href=\"index.php?act=findpost&pid=159011\"][{POST_SNAPBACK}][/a]

I am more likely to trust reviews with actual photos than a listing of facts, figures, and speculation in this photo forum :-)

Anyway, I'll have the camera and decide for myself soon enough.
Logged
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #26 on: December 07, 2007, 01:52:51 PM »
ReplyReply

Quote
I am more likely to trust reviews with actual photos than a listing of facts, figures, and speculation in this photo forum :-)
[{POST_SNAPBACK}][/a]

Here's a 14-bit RAW from a 1Dmk3 at ISO 100.  This is the lowest 256 RAW levels, untouched, and quantized to 12, 11, and 10 bits.  This is dark brown, almost black hair pushed to about ISO 2800 to 3200.  All quantized, then white-balanced and interpolated to fill the RGB of all pixels in the RAW (simple demosaicing). All have an extra 7 bits of precision.  As you can see, the difference between 14-bit and 12-bit is miniscule, and this is the camera/ISO with the lowest read noise of any 14-bit DSLR.  My estimation is that when the minimum (read or blackframe) noise is about 1.3 to 1.4 RAW levels (ADUs), the bit depth is optimum.  For this image, the read noise is 1.22 12-bit ADUs, or 4.88 14-bit ADUs.

Blue channel only:

[a href=\"http://www.pbase.com/jps_photo/image/84815836/original]http://www.pbase.com/jps_photo/image/84815836/original[/url]

In color:

http://www.pbase.com/jps_photo/image/76001165/original
« Last Edit: December 07, 2007, 04:11:12 PM by John Sheehy » Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #27 on: December 07, 2007, 07:41:21 PM »
ReplyReply

Quote
Reviewers who discuss the benefits of HTP are effectively pointing out how image processing modifications combined with a higher bit-depth in the RAW image data can improve the rendition of highlight details compared to earlier cameras without those features

Those reviewers are repeating slogans instead of going into details.

Quote
the 14-bit A/D conversion enhances the effect of HTP because it provides 4 times as many gradations to choose from as the RAW image data from cameras with 12-bit A/D conversion

This too is worthless blathering. One needs to think about what effect exactly the greater bit depth does for HTP. It's easy to show, that it does nothing.

I hoped that the increased bit depth has some visible effect on the images. I am still not ready to accept the quite amateurish "proofs", that the extra bits are only noise.

I invested inordinate amount of time with shooting scenes designed to demonstrate, that the two extra bits do in fact represent graduations and with analyzing the results. Several times I though I had it, but after reviewing, I alway came to the conclusion that my proofs were questionable.

However, independently of the worthiness of these "new" bits, there is an other side of the issue, namely when and how the intermediate levels can be utilized. If at all, then only in situations, when the low end gets boosted several stops or the contrast at the low end gets highly (but very highly) increased. Neither of these are happening with HTP (note, that the advantage of HTP is in recording JPEG).
Logged

Gabor
203
Guest
« Reply #28 on: December 08, 2007, 08:55:22 AM »
ReplyReply

Quote
Those reviewers are repeating slogans instead of going into details.
This too is worthless blathering. One needs to think about what effect exactly the greater bit depth does for HTP. It's easy to show, that it does nothing.

[{POST_SNAPBACK}][/a]

Speaking of worthless blathering, would you (or John) care to demonstrate your findings here? I'd like to see your proof that what Canon has incorporated into these cameras does nothing useful. All the techno-speak does very little for me. Let see some real evidence, rather than just words.  

In the mean time, here are two of the places where I saw stuff which seems to suggest that there are demonstrable improvements.

[a href=\"http://www.bobatkins.com/photography/digital/canon_eos_40D_review_6.html]http://www.bobatkins.com/photography/digit...D_review_6.html[/url]

http://www.dpreview.com/reviews/canoneos40d/page20.asp

Again, let's see proof guys, not just tech talk.
Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #29 on: December 08, 2007, 10:55:18 AM »
ReplyReply

Quote
Speaking of worthless blathering, would you (or John) care to demonstrate your findings here? I'd like to see your proof that what Canon has incorporated into these cameras does nothing useful. All the techno-speak does very little for me. Let see some real evidence, rather than just words

1. If there is a claim, that the camera has some improvements, then that claim needs to be proven, not the opposite.

2. I have not stated, that it does nothing useful. Just the opposite: I stated on another thread, that HTP is useful for JPEG. Its usefulness for raw files is very limited: the in-camera display is brighter than it would be with 1 stop underexposure, and the embedded JPEG file as well as the thumbnail are better with HTP. One may find it an advantage of HTP, that DPP and ACR automatically corrects the exposure by +1EV, though I don't like such covert actions. DPP adjusts the contrast in the highlights as well, like the in-camera conversion.

3. The 40D does have improvements (although it does not have "better colors" in raw). For me the difference between the 20D and the 40D was enough to but it (now, what to do with the former? I think I will make an IR camera from the 20D).

Furthermore, I am still not fully convinced, that the increased bit depth is really only noise. However, that needs to be proven, and I have not seen any proof of that, and I was unable to prove it convincingly (although I found some indications, that it is not or not always noise).

The review you cited from Bob Atkinson is an excellent demontration to the shallow blathering. the similar function in the EOS 1D MkIII is said to provide up to an extra stop of dynamic range in the highlight regions is an extraordinary nonsense (although he expresses this cautiously, "is said to"). This part of the review is restricted to guess, might, would.

DPReview did not go into details about the exact effect of HTP on raw data.
Logged

Gabor
N Walker
Sr. Member
****
Offline Offline

Posts: 295


WWW
« Reply #30 on: December 08, 2007, 11:28:22 AM »
ReplyReply

Quote
Chuck appears on other forums I've been on, and very tight-lipped about anything going on under the hood - you'll get much more useful answers by reverse-engineering the camera's output.
[{POST_SNAPBACK}][/a]


John,
I posed a similar question on the Adobe forums regarding the value of HTR using RAW - see Thomas Knoll's reply [a href=\"http://www.adobeforums.com/webx?14@@.3bb6a869.3c0330d0/0]http://www.adobeforums.com/webx?14@@.3bb6a869.3c0330d0/0[/url]

It is useful for jpeg's. In RAW all that is occurring is underexposure by 1 stop.  He will have reversed engineered the CR2 files to discover this.

Nick Walker
« Last Edit: December 08, 2007, 01:47:42 PM by Nick Walker » Logged

Slobodan Blagojevic
Sr. Member
****
Online Online

Posts: 5480



WWW
« Reply #31 on: December 08, 2007, 01:52:03 PM »
ReplyReply

Quote
... I'd like to see your proof that what Canon has incorporated into these cameras does nothing useful...

Nobody has ever said HTP "does nothing useful" in JPEG. Where we differ is RAW.

Perhaps we are differing in semantics here, so lets get that out of the way first. When people say (myself included) that HTP has no effect on RAW, they are actually saying: "HTP has no effect on RAW other than underexposing it by one stop".

Both sides further agree that HTP does affect how a RAW file is displayed on the camera's screen and later on in a RAW converter. Where both sides of the debate really differ is in how an HTP RAW shot affects highlights. My side of the debate says: HTP does not affect highlights any better then underexposing RAW file by one stop does.

The other side of the debate (i.e, Bob Atkins and very few others) seem to believe that HTP does more than that, i.e., that it somehow manages to preserve highlight detail better (i.e., better than underexposure does). If that would be true, i.e., preserving more highlight details, that would simply mean that we would have a one stop better dynamic range of the sensor. Even Canon does not claim that.

But let us suppose for a moment it is true (i.e., that HTP does increase the DR of the sensor). Why would then Canon offer cameras with HTP as an option only? Why would then anyone choose to cripple the sensor and blow highlights by not using HTP?

And this is why the belief in the HTP RAW "virtues" is dangerous: it makes people keep HTP on all the time (and why not, they say, who would not want a better DR?). By keeping it on all the time, they are underexposing every shot they take, even those not in danger of highlights being blown. Instead of "exposing to the right" they are consistently doing the opposite: exposing to the left, i.e., losing details and adding noise in mid-tones and shadows.
Logged

Slobodan

Flickr
500px
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #32 on: December 08, 2007, 02:05:25 PM »
ReplyReply

Quote
I posed a similar question on the Adobe forums regarding the value of HTR using RAW - see Thomas Knoll's reply http://www.adobeforums.com/webx?14@@.3bb6a869.3c0330d0/0

That reply is incorrect (sloppy). Not the exposure gets reduced but the ISO. This is the reason, why HTP is not available with ISO 100. It does not need any reverse engineering to find it out.
« Last Edit: December 08, 2007, 02:06:28 PM by Panopeeper » Logged

Gabor
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1273



WWW
« Reply #33 on: December 08, 2007, 02:54:03 PM »
ReplyReply

The more I read and analyse about the newly come 14-bit cameras, the more I think Canon and Nikon reached some agreement with all the manufacturers of memory cards, hard disks and storage devices in general. We are starting to deal with 25MB RAW files per photograph, that's beginning to be insane.

Lots of Mpx is probably fine for certain applications (large printings basically), but I am starting to question myself if going beyond 12 bit RAW files is of any use to improve image quality. At least meanwhile noise remains being the limiting factor to DR.
« Last Edit: December 08, 2007, 02:56:17 PM by GLuijk » Logged

Ray
Sr. Member
****
Offline Offline

Posts: 8845


« Reply #34 on: December 08, 2007, 02:56:51 PM »
ReplyReply

Quote
That reply is incorrect (sloppy). Not the exposure gets reduced but the ISO. This is the reason, why HTP is not available with ISO 100. It does not need any reverse engineering to find it out.
[a href=\"index.php?act=findpost&pid=159302\"][{POST_SNAPBACK}][/a]

I think the confusion here is the idea that HTP at say ISO 200 is the same as one stop underexposed at ISO 100. That implies, or could imply depending on how you interpret it, that using HTP with a particular exposure at ISO 200 is the same as dropping down to ISO 100 and dialing in -1 EV, when in fact it's the same as dialing in -1EV at ISO 200.

Whether the underexposure is acheived by dialing in the -1EV at ISO 200, or by the camera dropping down to ISO 100 but using the same exposure that has been auto-metered at ISO 200, is immaterial. The same number of photons arrive at the sensor, except, and it appears to be a big exception, higher ISO's can handle underexposure better, at least in Canon DSLRs.

I've verified this myself with the 5D. An image at ISO 100, say 1 stop underexposed, will have more shadow noise than it will if the ISO is bumped up to ISO 200, keeping the exposure the same. This apparently, (I just read what makes sense to me) is because at ISOs above base ISO, the analog signal is boosted prior to A/D conversion resulting in lower read noise. Same thing applies at all the other ISO's except ISO 3200. Comparing an image that's been underexposed by 4 stops at ISO 100 with the same image at literally the same exposure but at ISO 1600, really brings out the differences in noise in the mid-tones, lower mid-tones and shadows.
Logged
N Walker
Sr. Member
****
Offline Offline

Posts: 295


WWW
« Reply #35 on: December 08, 2007, 03:51:24 PM »
ReplyReply

Quote
That reply is incorrect (sloppy). Not the exposure gets reduced but the ISO. This is the reason, why HTP is not available with ISO 100. It does not need any reverse engineering to find it out.
[a href=\"index.php?act=findpost&pid=159302\"][{POST_SNAPBACK}][/a]

Your reply is (sloppy) I am fully aware that this is the reason why only 200 ISO is available, it doesn't take a rocket scientist to work this out, but the end result is 1 stop of underexposure which ever way you look at it.
Logged

Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #36 on: December 08, 2007, 07:37:45 PM »
ReplyReply

Quote
Whether the underexposure is acheived by dialing in the -1EV at ISO 200, or by the camera dropping down to ISO 100 but using the same exposure that has been auto-metered at ISO 200, is immaterial

You are over-complicating this very simple issue. It has nothing but absolute nothing to do with metering, EV bias, etc.

The camera (at least my cameras) displays the shutter speed, the aperture (and only these determine the exposure, not ISO), as well as the ISO. When shooting, that very exposure will be used, i.e. the indicated shutter and aperture, and the ISO gets "manipulated".

It is not subject to interpretation, that HTP leaves the exposure unchanged but reduces the ISO.
Logged

Gabor
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #37 on: December 08, 2007, 07:40:30 PM »
ReplyReply

Quote
Your reply is (sloppy) I am fully aware that this is the reason why only 200 ISO is available, it doesn't take a rocket scientist to work this out, but the end result is 1 stop of underexposure which ever way you look at it.

IMO you need to study the difference between having to say something and having something to say.
Logged

Gabor
Ray
Sr. Member
****
Offline Offline

Posts: 8845


« Reply #38 on: December 08, 2007, 11:45:42 PM »
ReplyReply

Quote
You are over-complicating this very simple issue. It has nothing but absolute nothing to do with metering, EV bias, etc.

The camera (at least my cameras) displays the shutter speed, the aperture (and only these determine the exposure, not ISO), as well as the ISO. When shooting, that very exposure will be used, i.e. the indicated shutter and aperture, and the ISO gets "manipulated".

It is not subject to interpretation, that HTP leaves the exposure unchanged but reduces the ISO.
[a href=\"index.php?act=findpost&pid=159350\"][{POST_SNAPBACK}][/a]

Everything is subject to interpretation, my friend, even the most objective scientific theories.

The confusion is in the language. I notice that John Sheehy has amended his explanations a couple of times.

So let me rephrase what I understand is going on here, and you can correct me if I'm wrong.

(1) Activating HTP does not change the actual and real exposure (shutter speed). It just drops the ISO setting down a stop. In relation to the 'hidden' lower ISO setting that the camera has surreptitiously used, the shot now has the equivalent of one stop less exposure. Whether or not the shot is actually underexposed as a result, will depend on other circumstance.

(2) If the shutter speed at the higher ISO is correct for a perfect ETTR (and one is shooting RAW), then activating HTP will be disadvantageous and produce more noise in the shadows.

(3) If the shutter speed at the higher ISO would have been correct for an ETTR if one had been shooting RAW but one happens to be shooting in JPEG mode, then one is protected from blown highlights by using HTP.

This is the purpose of HTP.
Logged
N Walker
Sr. Member
****
Offline Offline

Posts: 295


WWW
« Reply #39 on: December 09, 2007, 05:07:50 AM »
ReplyReply

Quote from: Ray,Dec 9 2007, 05:45 AM
Everything is subject to interpretation, my friend, even the most objective scientific theories.

The confusion is in the language. I notice that John Sheehy has amended his explanations a couple of times.

Quote from: Ray,Dec 9 2007, 05:45 AM
Everything is subject to interpretation, my friend, even the most objective scientific theories.

The confusion is in the language. I notice that John Sheehy has amended his explanations a couple of times.

Ray,

There is no confusion in the language as HTP results in one stop less of exposure at the time of recording the exposure. The shutter speed and aperture selected are not altered by HTP otherwise it wouldn't work using manual shutter and aperture speeds. Therefore the only means left available for HTP to work is by dropping the sensor to 100 ISO to reduce the exposure. To meet Canon's HTP feature demands a minimum setting of 200 ISO to enable this to happen.

In the case of Jpegs the Digic III image processor adjusts the underexposed RAW data and processes the image accordingly, hence Canon's warning of possible increased noise.

You also said -
"IMO you need to study the difference between having to say something and having something to say."

Maybe you should reflect in the manner in which you replied to my first post. Your choice of words, and the wrong assumption that I was not aware that the sensor dropped to 100 ISO. You could have asked me if I was aware of this? This information has been common knowledge to me, and other photographers, for several months, as a result of seeking assistance on this matter.

I now move on.
« Last Edit: December 09, 2007, 05:23:01 AM by Nick Walker » Logged

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

Ad
Ad
Ad