Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Doug Peterson's article gives good insight into development of the IQ250  (Read 6789 times)
ErikKaffehr
Sr. Member
****
Online Online

Posts: 7871


WWW
« on: February 08, 2014, 01:54:06 AM »
ReplyReply

Hi,

Doug Peterson posted a nice article about the development of the IQ-250 in special but also the IQ series in general: http://www.luminous-landscape.com/essays/the_phase_one_iq250_cmos_fully_realized.shtml

I feel it is a fascinating/good read. It also gives some insights into colour handling at Phase One. Interesting read for geeks like me.

Good info by Doug as usual, much appreciated.

Best regards
Erik

Logged

Christoph C. Feldhaim
Sr. Member
****
Offline Offline

Posts: 2509


There is no rule! No - wait ...


« Reply #1 on: February 08, 2014, 05:02:44 AM »
ReplyReply

I found especially interesting learning about the conflict between color accuracy and speed in the design of the CFAs.
Cheers
~Chris
Logged

ErikKaffehr
Sr. Member
****
Online Online

Posts: 7871


WWW
« Reply #2 on: February 08, 2014, 06:40:06 AM »
ReplyReply

Hi,

It is much about how selective a CFA is. This article gives some insight: http://www.dxomark.com/Reviews/Canon-500D-T1i-vs.-Nikon-D5000/Color-blindness-sensor-quality

See the figures below canon has a more permissive "red" channel, so it will permit more light, but needs more math calculate colour.

I wouldn't be sure that very "orthogonal" CFA-s are optimum, if there is overlap between RGB pixels an oversaturated look may result.

Canon 500D-T1iNikon-D5000


I found especially interesting learning about the conflict between color accuracy and speed in the design of the CFAs.
Cheers
~Chris
« Last Edit: February 08, 2014, 06:55:13 AM by ErikKaffehr » Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2855



« Reply #3 on: February 08, 2014, 06:48:19 AM »
ReplyReply

I found especially interesting learning about the conflict between color accuracy and speed in the design of the CFAs.
Cheers
~Chris

That is not new information. DXO published an interesting review on that subject some time ago and gave a more detailed explanation. Data on the channel decomposition and color matrix of the new camera will be interesting.

Bill
Logged
Christoph C. Feldhaim
Sr. Member
****
Offline Offline

Posts: 2509


There is no rule! No - wait ...


« Reply #4 on: February 08, 2014, 09:56:52 AM »
ReplyReply

Just for comparison - here we have the human eye:



Quite a difference, isn't it?
Cheers
~Chris
Logged

jerome_m
Sr. Member
****
Offline Offline

Posts: 651


« Reply #5 on: February 08, 2014, 10:19:06 AM »
ReplyReply

See the figures below canon has a more permissive "red" channel, so it will permit more light, but needs more math calculate colour.

I have read that because Canon filters are less discriminant between red and green, they give more pleasing results for skin tones under poor light, e.g. fluorescent lights used in sports venues or at home. Conversely, MF cameras are more discriminant and are more sensitive to poor light, but this is not a problem for their intended use. Indeed if one tries to take a portrait under typical home lights, 24*36 cameras will give more pleasing results than a MF, even when one tries to correct white balance (you can try it yourself). On the other hand, MF cameras are better at showing differences between orange and red.

As evidenced by the curves above, the human eye does not discriminate very well between red and green. A camera which discriminates better than the human eye will record colour differences that we do not normally see. That will look unnatural.

Also: more permissive filters let more signal in, but they need stronger colour processing to compensate. That stronger processing increases noise, so I am not sure that there is much gain on the noise front.
« Last Edit: February 08, 2014, 10:25:28 AM by jerome_m » Logged
Christoph C. Feldhaim
Sr. Member
****
Offline Offline

Posts: 2509


There is no rule! No - wait ...


« Reply #6 on: February 08, 2014, 10:24:39 AM »
ReplyReply

...As evidenced by the curves above, the human eye does not discriminate very well between red and green. A camera which discriminates better than the human eye will record colour differences that we do not normally see. That will look unnatural.

Exactly!
Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2855



« Reply #7 on: February 08, 2014, 11:24:19 AM »
ReplyReply

I have read that because Canon filters are less discriminant between red and green, they give more pleasing results for skin tones under poor light, e.g. fluorescent lights used in sports venues or at home. Conversely, MF cameras are more discriminant and are more sensitive to poor light, but this is not a problem for their intended use. Indeed if one tries to take a portrait under typical home lights, 24*36 cameras will give more pleasing results than a MF, even when one tries to correct white balance (you can try it yourself). On the other hand, MF cameras are better at showing differences between orange and red.

That is an interesting observation regarding channel decompostion of Nikon vs Canon. However, the assertion that MF is more discriminant is doubtful. It would depend on the CFA filters and these could vary among different MF and 24x36 sensors. The DXO data for the Nikon D800e and Phase One IQ180 demonstrate that the D800 is more discriminant than the IQ180. How the new Phase One would compare is not yet available.

Bill

Logged
jerome_m
Sr. Member
****
Offline Offline

Posts: 651


« Reply #8 on: February 08, 2014, 12:02:10 PM »
ReplyReply

However, the assertion that MF is more discriminant is doubtful.

I meant: "typical MF cameras are more discriminant between red and green than typical Canon cameras".

The message was that the characteristics of the CFA are a compromise between a better differentiation between orange and red (which may be useful for skin tones) and more consistent rendering under some fluorescent lights (which may be useful for a camera intended for sports or for home shooting). It is not only the high iso performance which is part of the equation.

For me, the idea that rendering under fluorescent lights may need to be optimised (or not) was an eye opener. When I shoot under poor light with my 24x36 camera, a simple white balance correction will make skin tones acceptable. If I do the same with a MF camera, getting something which does not look like a zombie attack is much more difficult.
« Last Edit: February 08, 2014, 12:09:52 PM by jerome_m » Logged
Telecaster
Sr. Member
****
Offline Offline

Posts: 911



« Reply #9 on: February 08, 2014, 01:21:56 PM »
ReplyReply

As evidenced by the curves above, the human eye does not discriminate very well between red and green.

Just imagine the processing that goes into creating the red/green tonalities we see! It's no wonder that ~30% of our grey-matter computing power is devoted to vision.

-Dave-
Logged
ErikKaffehr
Sr. Member
****
Online Online

Posts: 7871


WWW
« Reply #10 on: February 08, 2014, 03:26:42 PM »
ReplyReply

Hi Bill,

Are you referring to SMI (Sensor Metamerism Indexrating)?

It just depends on the 16 coloured fields of the ColorChecker.

I did test my P45+ on the colour checker, and had a decent match using the Adobe Standard Profile, albeit not even close to my Sony Alpha 99 SLT, but I got bad colours on "red" flower which got rendered 'red violet'. A new profile fixed the issue. The point is that SMI would be better for the Adobe Standard Profile but I definitively preferred my own profile.



Best regards
Erik


That is an interesting observation regarding channel decompostion of Nikon vs Canon. However, the assertion that MF is more discriminant is doubtful. It would depend on the CFA filters and these could vary among different MF and 24x36 sensors. The DXO data for the Nikon D800e and Phase One IQ180 demonstrate that the D800 is more discriminant than the IQ180. How the new Phase One would compare is not yet available.

Bill


« Last Edit: February 08, 2014, 03:29:35 PM by ErikKaffehr » Logged

yaredna
Jr. Member
**
Offline Offline

Posts: 51


« Reply #11 on: February 08, 2014, 08:34:21 PM »
ReplyReply

Great article, likely written by two people: the first paragraph (foreword) has a style totally different than the body of the article. Informal vs formal.

That said, The article is very informative, who ever wrote it. Kudos to Doug for sharing!

NY
Logged
Wayne Fox
Sr. Member
****
Offline Offline

Posts: 2925



WWW
« Reply #12 on: February 08, 2014, 10:12:17 PM »
ReplyReply

Great article, likely written by two people: the first paragraph (foreword) has a style totally different than the body of the article. Informal vs formal.

That said, The article is very informative, who ever wrote it. Kudos to Doug for sharing!

NY

are you talking about the forward vs the "article".  Certainly the article has seen considerable work, and perhaps has been written over a period of time.  The forward seems to be added to alert viewers as to what to except in the article.  But I don't know they were written by different people, just different objectives for something written probably at different times.  I get the sense the forward was added for the purpose of posting to this site, the article itself will probably see a broader distribution.

An interesting read...
Logged

ErikKaffehr
Sr. Member
****
Online Online

Posts: 7871


WWW
« Reply #13 on: February 09, 2014, 01:09:12 AM »
ReplyReply

Hi,

The figure below shows colour interpretation as measured by Imatest on the P45+ using Capture 1 (7.1.3). It shows that greens are shifted towards yellows, and this is very visible in landscape shots:


The image below shows my A99-SLT with a profile made by Adobe DNG Profile editor:


I know that Digital Transitions has a department  working on repro-photography, and I recall reading they could get DeltaE around 2.0 with an IQ 280 (?) on the test chart they use. Accurate is not pleasant, but reproducing the sixteen colours that happen to be to the standard reference point in the industry is a good starting point. Those colours were chosen for a reason.

On the other hand, Adobe Standard profile yielded Delta E = 5.65, a lot better than Capture One, but the Adobe standard profile was not acceptable to me on the P45+.

One of the most important aspects is white balance, the images below are identical, and the same surface was used for WB, but on left one the sunlit part was used while on the right one the shadow part was used for WB. I guess that getting WB right is extremely important for colour rendition.
Best regards
Erik
« Last Edit: February 09, 2014, 01:11:16 AM by ErikKaffehr » Logged

Christoph C. Feldhaim
Sr. Member
****
Offline Offline

Posts: 2509


There is no rule! No - wait ...


« Reply #14 on: February 09, 2014, 02:00:26 AM »
ReplyReply

...
I guess that getting WB right is extremely important for colour rendition.
...

+1000
Logged

Kolor-Pikker
Jr. Member
**
Offline Offline

Posts: 57


« Reply #15 on: February 09, 2014, 05:06:41 AM »
ReplyReply

I should add that speed isn't the only reason for reduced color accuracy in most DSLRs and consumer cameras, by intentionally manipulating the color response, it's possible to make images look better under poor lighting conditions, which tend to have "spikes" in the color spectrum.
To put it another way, if you take a shot of a person under fluorescent available light on medium format, they'll likely come out looking like a corpse, but with a DLSR the result is instead passable because these are the conditions it was tuned to shoot in. Canon for instance takes a great deal of pride in it's sports photography, so it only makes sense that they would want to tune the sensor for a lack of red and a spike in the blue-green spectrum. High color acuity would actually be detrimental to the average camera user.

Also in the SAS section, second paragraph "These usability advantages has been attractive to", supposed to be "have been".

Hope they really do make CMOS sing.
« Last Edit: February 09, 2014, 05:09:36 AM by Kolor-Pikker » Logged
jerome_m
Sr. Member
****
Offline Offline

Posts: 651


« Reply #16 on: February 09, 2014, 05:21:01 AM »
ReplyReply

I should add that speed isn't the only reason for reduced color accuracy in most DSLRs and consumer cameras, by intentionally manipulating the color response, it's possible to make images look better under poor lighting conditions

This is exactly what I tried to say yesterday, but your text is better.
Logged
jerome_m
Sr. Member
****
Offline Offline

Posts: 651


« Reply #17 on: February 09, 2014, 05:38:06 AM »
ReplyReply

The figure below shows colour interpretation as measured by Imatest on the P45+ using Capture 1 (7.1.3). It shows that greens are shifted towards yellows, and this is very visible in landscape shots:


True, but this is not what we were talking about. This is about a given patch of pigment being rendered of a slightly different colour than the one the human eye would perceive. Metamerism is about two given patches of different pigments which appear of the same colour to the human eye but are rendered of a different colour by the camera.

The effect is common with certain flowers. They appear to be of the same colour to the human eye (e.g. both are red) but are of different colour when photographed (e.g. one is red and one is orange).

Now, imagine a portrait situation. For an healthy model, patches of skin with or without underlying veins appear to be almost of the same colour to the human eye. Our eyes have evolved to perceive skin in that way. Cosmetics appear to our eyes to be of skin colour, because they are designed in that way. However, the spectral content of reflected light may be slightly different for skin, skin with veins and cosmetic pigments. It is just that our eyes are not sensitive to these spectral differences. But if the camera is sensitive to these differences, we will see them on the photographs. In truth, the camera is more sensitive to the differences than the human eye, but the results are still not what we want, for example because veins are more visible.
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3863


« Reply #18 on: February 09, 2014, 06:41:20 AM »
ReplyReply

True, but this is not what we were talking about. This is about a given patch of pigment being rendered of a slightly different colour than the one the human eye would perceive. Metamerism is about two given patches of different pigments which appear of the same colour to the human eye but are rendered of a different colour by the camera.

Hi Jerome,

You've read my mind, it's exactly what I was thinking. What the Imatest charts show it how a profile affects/shifts colors, not how well one can discriminate between similar colors, or even metameres (which by definition will have difficulty to render correctly when changing white-balance).

The Color Sensitivity Index and Sensitivity Metamerism Index are metrics for the discrimination sensitivity, not color accuracy. A high color sensitivity offers opportunities to very precicely tweak certain color differences, either making them more pronounced, or less obvious.

Cheers,
Bart
Logged
Kolor-Pikker
Jr. Member
**
Offline Offline

Posts: 57


« Reply #19 on: February 09, 2014, 06:48:17 AM »
ReplyReply

True, but this is not what we were talking about. This is about a given patch of pigment being rendered of a slightly different colour than the one the human eye would perceive. Metamerism is about two given patches of different pigments which appear of the same colour to the human eye but are rendered of a different colour by the camera.

The effect is common with certain flowers. They appear to be of the same colour to the human eye (e.g. both are red) but are of different colour when photographed (e.g. one is red and one is orange).

Now, imagine a portrait situation. For an healthy model, patches of skin with or without underlying veins appear to be almost of the same colour to the human eye. Our eyes have evolved to perceive skin in that way. Cosmetics appear to our eyes to be of skin colour, because they are designed in that way. However, the spectral content of reflected light may be slightly different for skin, skin with veins and cosmetic pigments. It is just that our eyes are not sensitive to these spectral differences. But if the camera is sensitive to these differences, we will see them on the photographs. In truth, the camera is more sensitive to the differences than the human eye, but the results are still not what we want, for example because veins are more visible.

This is pretty deep topic, ultimately the colors you get are a combination of the light hitting the subject, the reflective properties of the subject, and the sensor's response. This is why the article talks about using charts only as as a starting point for profiling, ink on paper is not perfectly representative of what you'll actually be shooting.

I recall reading an article written by a director of photography who was shooting a film and his experience with lighting and the resulting color, at one point he wrote about how he wanted to mix in a litepanel (a type LED light used for video) to add extra illumination the face of an actress, but when the film was scanned, the face if the actress was... magenta! It just wasn't possible to fix the color in post either, because as an engineer explained to him "if the color isn't there to begin with, it can't be brought back in post", this was the result of the LED spectrum, the reflective properties of the skin/makeup and the color response of film matching up in the worst way possible.

Different types of illumination produce different amounts of light in different parts of the spectrum, if a light doesn't produce blue, like low pressure sodium for instance, no amount of blue reflectance and sensitivity will bring it back.

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

Ad
Ad
Ad