Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Focusing in the Digital Era  (Read 10231 times)
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« on: October 24, 2006, 11:54:00 PM »
ReplyReply

This is a comment on the article 'Focusing in the Digital Era - Part I' recently published on the LL site.

I am not going to argue here about the details and findings in the article. What I want to comment on is the author's basic premise that in the case of digital capture, degree of enlargement, defined as the percentage of increase of the sensor dimensions to produce a specific sized print, is of relevance. Assuming this premise results, for example, in the reasoning that an image captured by a (theoretically perfect) sensor smaller but of equal MP number than a larger (theoretically perfect) sensor can 'support' less 'enlargement', just because of its smaller geometric dimensions.

I find it hard to agree with this assertion, although it is a very common one in digital photography fora. A digital image, I believe, is a 'virtual' entity. It does not have dimensions per se, the way a piece of film has. By the time it is captured it exists in storage and it is independent of the geometric dimensions of the capture sensor. I can see no way of attributing absolute geometric dimensions to this image which are then somehow 'enlarged' to produce a print.

'Enlargement' and 'enlargeability' in the digital era depends on MP content, information amount so to speak. The degree of enlargeability is directly related to the amount of information contained and not on any non-existent geometric dimensions of the captured image.

So, can't we draw a parallel between digital and analogue capture with regards to 'enlargeability'? I believe we can but we have to go one level up in the film case to draw parallels. Examine why is that a larger piece of film can support more enlargement than a smaller one.

Information content as a measure of enlargeability is true for the film case as well. But in that case, assuming the use of the same film emulsion with a set ability for information capture per unit of area, a larger piece of film directly means more information storage capability.

Surely there are quantities directly related to the geometric dimensions and MP number of a capture sensor that will affect the DoF (Circle of Confusion), pixel pitch for one, but the author seems to ignore these. In fact, I have yet to see a conscise and convincing essay on the net dealing with the definition of the, largely subjective, CoC using terms applicable to the digital era.

I suppose this is a difference in semantics but a very important one I believe.

I would welcome constructive arguments for or against my premise.

Nikos Razis
Athens, Greece
« Last Edit: October 25, 2006, 02:18:39 AM by NikosR » Logged

Nikos
Nemo
Sr. Member
****
Offline Offline

Posts: 276



« Reply #1 on: October 25, 2006, 03:38:24 AM »
ReplyReply

Ferguson's analysis is very interesting, and a pleasure to read, but it shows an obvious thing: the focus plane is just a plane, not a gross wall.

The 135 marks show the limits for perceived sharpness in the print, and this depends on the print size (8'x12' was assumed). 100% crops can not show this. The sharpest crop and the blurrest crop will not be so different in the print if it is sufficiently small. Of course, they are different, but the human eye cannot perceive the difference. This is the key point.

However, 100% crops show the differences that we cannot see on paper. The same was the case with film. There is nothing special about digital cameras here.

Ferguson explains, and this is true, that in the digital era we tend to make bigger prints. He is also right in pointing to the outdated conventions of the 30s for the circle of confusion. Actually we know that the human eye can perceive details much smaller than 30 microns.
« Last Edit: October 25, 2006, 08:30:50 AM by Nemo » Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #2 on: October 25, 2006, 03:44:42 AM »
ReplyReply

Quote
Ferguson's analysis is very interesting, and a pleasure to read, but it shows an obvious thing: the focus plane is just a plane, not a gross wall.

The 135 marks show the limits for perceived sharpness in the print, and this depends on the print size. 100% crops can not show this. The sharpest crop and the blurest crop will not be so different in the print if it is sufficiently small. Of course, they are different, but the human eye cannot perceive the difference. This is the key point.

However, 100% crops show the differences that we cannot see on paper. The same was the case with film. There is nothing special about digital cameras here.

Ferguson explains, and this is true, that in the digital era we tend to make bigger prints. He is also right in pointing to the outdated conventions of the 30s for the circle of confusion. Actually we know that the human eye can perceive details much smaller than 30 microns.
[a href=\"index.php?act=findpost&pid=82153\"][{POST_SNAPBACK}][/a]


The author is right in many things he says in the article. He is also quite accurate in many of his subjective comments. My problem is his attempt at a theoretical substantiation of his observations.

I made the original post just to pin-point an incorrect in my opinion assertion about 'enlargeability' that is made in the beginning of the article, and repeated unsubstantiated in other places in the text.
Logged

Nikos
madmanchan
Sr. Member
****
Offline Offline

Posts: 2110


« Reply #3 on: October 25, 2006, 06:22:13 AM »
ReplyReply

I think one of the main points that Gary is trying to make is that unless we are careful about how we focus, then the bottleneck on image quality will be the limited depth of field imparted by the lens. It doesn't matter how many pixels the sensor has. Imagine using a 50 mm lens on a camera. No matter what kind of sensor the camera has or how many pixels it has, the lens will form (i.e., project) the same image onto the sensor plane (i.e., the imaging plane). The projected image depends only on the optics of the lens, how the lens is focused, and the laws of physics, nothing else.

Suppose there is a blurry spot in the resulting image that measures 2 mm by 2 mm. It doesn't matter whether this spot is digitized by the sensor using 50 pixels or 5000 pixels. The total amount of "image information" within that 2x2 mm area is constant. If that 2x2 mm area is enlarged in a print to be 20x20 mm, it will look equally blurry in either case.

To carry this example one step further: the only way in which 5000 pixels will be more useful than 50 pixels is if that "blurry spot" described above is actually sharp enough so that 5000 pixels can pick up detail that 50 pixels cannot. (In semi-math speak, the frequency content of that area has to be sufficiently high, otherwise adding more samples is useless.) And the only way the blurry spot can be "sharp enough" is by making sure that the area lies enough within the depth of field.

Eric
Logged

NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #4 on: October 25, 2006, 06:47:49 AM »
ReplyReply

Quote
I think one of the main points that Gary is trying to make is that unless we are careful about how we focus, then the bottleneck on image quality will be the limited depth of field imparted by the lens. It doesn't matter how many pixels the sensor has. Imagine using a 50 mm lens on a camera. No matter what kind of sensor the camera has or how many pixels it has, the lens will form (i.e., project) the same image onto the sensor plane (i.e., the imaging plane). The projected image depends only on the optics of the lens, how the lens is focused, and the laws of physics, nothing else.

Suppose there is a blurry spot in the resulting image that measures 2 mm by 2 mm. It doesn't matter whether this spot is digitized by the sensor using 50 pixels or 5000 pixels. The total amount of "image information" within that 2x2 mm area is constant. If that 2x2 mm area is enlarged in a print to be 20x20 mm, it will look equally blurry in either case.

To carry this example one step further: the only way in which 5000 pixels will be more useful than 50 pixels is if that "blurry spot" described above is actually sharp enough so that 5000 pixels can pick up detail that 50 pixels cannot. (In semi-math speak, the frequency content of that area has to be sufficiently high, otherwise adding more samples is useless.) And the only way the blurry spot can be "sharp enough" is by making sure that the area lies enough within the depth of field.

Eric
[a href=\"index.php?act=findpost&pid=82165\"][{POST_SNAPBACK}][/a]

Most of what  you're saying seem true to me. Information content is only useful data and a blurry patch will be blurry regardless with how many pixels it is digitized.

However, where is the direct connection between sensor dimensions and what you are describing? Why would a larger sized sensor be any different? (leaving lens resolution and diffraction issues aside).

The only connection I can establish is, as I hinted in my OP, pixel pitch.

Remember, my gripe wih regards to the article is the explicit and implied direct connection between sensor size and ability to enlarge. Not wanting to tread onto the treacherous path of CoC determination and DoF calculation for digital capture, I would believe that a smaller sized sensor (hence with higher pixel pitch) would provide deeper DoF than a larger one of the same Mp count. Not the other way around as the article seems to imply.
Logged

Nikos
Caracalla
Full Member
***
Offline Offline

Posts: 156



WWW
« Reply #5 on: October 25, 2006, 07:44:04 AM »
ReplyReply

I found the article rather convenient & straightforward, although not entirely comprehensive. Waiting for the PART II, PART III.

Regards
« Last Edit: October 25, 2006, 11:45:01 AM by Caracalla » Logged
bjanes
Sr. Member
****
Online Online

Posts: 2825



« Reply #6 on: October 25, 2006, 07:45:43 AM »
ReplyReply

Quote
Most of what  you're saying seem true to me. Information content is only useful data and a blurry patch will be blurry regardless with how many pixels it is digitized.

However, where is the direct connection between sensor dimensions and what you are describing? Why would a larger sized sensor be any different? (leaving lens resolution and diffraction issues aside).

The only connection I can establish is, as I hinted in my OP, pixel pitch.

Remember, my gripe wih regards to the article is the explicit and implied direct connection between sensor size and ability to enlarge. Not wanting to tread onto the treacherous path of CoC determination and DoF calculation for digital capture, I would believe that a smaller sized sensor (hence with higher pixel pitch) would provide deeper DoF than a larger one of the same Mp count. Not the other way around as the article seems to imply.
[{POST_SNAPBACK}][/a]

There is no need to deal with the complexities of CoC and DoF calculations since these caclulations can be done by online calculators such as the one on Cambridge in Color:

[a href=\"http://www.cambridgeincolour.com/tutorials/DOF-calculator.htm]http://www.cambridgeincolour.com/tutorials...-calculator.htm[/url]

If you plug in the variables for a 35 mm full frame with a 50mm lens and 6 by 7 cm with an 80 mm lens, you will see that the smaller format has a greater depth of field.

The resolution of the sensor does not enter directly into the calculation, but if the increased details associated with great depth of field can not be resolved by the sensor, the full depth of field can not be utilized.

Bill
Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #7 on: October 25, 2006, 07:53:06 AM »
ReplyReply

Quote
There is no need to deal with the complexities of CoC and DoF calculations since these caclulations can be done by online calculators such as the one on Cambridge in Color:

http://www.cambridgeincolour.com/tutorials...-calculator.htm

If you plug in the variables for a 35 mm full frame with a 50mm lens and 6 by 7 cm with an 80 mm lens, you will see that the smaller format has a greater depth of field.

The resolution of the sensor does not enter directly into the calculation, but if the increased details associated with great depth of field can not be resolved by the sensor, the full depth of field can not be utilized.

Bill
[a href=\"index.php?act=findpost&pid=82173\"][{POST_SNAPBACK}][/a]

I'd rather limit the discussion of DoF for digital sensors, since I believe it is a bit off-topic in this thread (with regards to my OP) and is a thorny issue IMO.

However, since you mention it, have you wondered how the author of that particular calculator has determined the CoC for the various digital sensors? I have seen no conscise treatment for CoC determination for digital sensors.

The author of the DoF calculator does not treat digital and film any differently, quoting the sensor dimensions enlargement required to bring it up to 8x10 as the way to determine the CoC to be used. But this assertion that a digital image is enlarged from the sensor size to the final print is what I'm arguing against.

Just to put it another way, although I suspect this statement will not go down well with many: It is my assertion that an APS-C sized, for example, sensor of 3Mp has a very different CoC, and hence DoF, than an APS-C sized sensor of 12Mp. This might sound strange at first thought, but think about it a little in terms of apparent sharpness of a print produced at a certain size.

In fact, your statement in your last sentence, supports well what I'm suggesting. I suggest the same thing, but without having to resort to notions of 'depth of field utilization'.
« Last Edit: October 25, 2006, 08:19:39 AM by NikosR » Logged

Nikos
russell a
Sr. Member
****
Offline Offline

Posts: 389


WWW
« Reply #8 on: October 25, 2006, 08:27:12 AM »
ReplyReply

My take-away from the article is that one should focus carefully.  Is this a revelation?  The biggest practical change between the film and digital era is this regard is that it is easier and cheaper to make larger prints and so the results may come under greater scrutiny ("stop that fellow with the loupe!").  On the other side, there are digital tools that aid upscaling.
Logged
howiesmith
Guest
« Reply #9 on: October 25, 2006, 08:28:12 AM »
ReplyReply

Quote
In fact, I have yet to see a conscise and convincing essay on the net dealing with the definition of the, largely subjective, CoC using terms applicable to the digital era.


[a href=\"index.php?act=findpost&pid=82136\"][{POST_SNAPBACK}][/a]


I think that is because "DOF doesn't care about the recording media type or size ... ."
Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #10 on: October 25, 2006, 08:31:49 AM »
ReplyReply

Quote
I think that is because "DOF doesn't care about the recording media type or size ... ."
[a href=\"index.php?act=findpost&pid=82181\"][{POST_SNAPBACK}][/a]

I beg to differ. These notions are inherent in the traditional calculations for DoF. The related parameter is the CoC which, while subjective to a large extent, has always been defined by convention in terms of film capture area and output size.
Logged

Nikos
Nemo
Sr. Member
****
Offline Offline

Posts: 276



« Reply #11 on: October 25, 2006, 08:36:11 AM »
ReplyReply

If fact, DoF depends on media size. It depends on the Circle of Confusion. The Circle of Confusion depends on the resolutive capacity of the observer's eyes, the film or sensor size and the print size.

DoF depends on other variables as well, like focal lengh, aperture, distance to the subject and hyperfocal distance.

http://en.wikipedia.org/wiki/Depth_of_fiel...f_field_formula

Digital cameras are under the same rules. The only difference is a constrain related to minimum CoC and pixel size.
Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #12 on: October 25, 2006, 08:50:27 AM »
ReplyReply

Quote
If fact, DoF depends on media size. It depends on the Circle of Confusion. The Circle of Confusion depends on the resolutive capacity of the observer's eyes, the film or sensor size and the print size.

As I have already exlained, I fail to see the direct relationship between sensor size and CoC. Asserting that a direct relationship exists (similar to the film case) would entail that one accepts the concept of 'enlarging' the sensor dimensions to the final output to be judged (say 8x10).

As I've repeatedly said, I fail to see this direct relationship in the case of digital capture, reason being the sensor is never enlarged. It just produces a digital representation of the image with only Mp resolution and as its 'native dimension'.

DoF equations, and in particular the conventions regarding the CoC need to be changed or updated to properly account for digital capture.
Logged

Nikos
howiesmith
Guest
« Reply #13 on: October 25, 2006, 09:12:26 AM »
ReplyReply

Quote
As I've repeatedly said, I fail to see this direct relationship in the case of digital capture, reason being the sensor is never enlarged. It just produces a digital representation of the image with only Mp resolution and as its 'native dimension'.

[a href=\"index.php?act=findpost&pid=82186\"][{POST_SNAPBACK}][/a]

Perhaps you fail to see a relationship because none exists.  Circle of confusion is an optical term with nothing to with cameras (digital or otherwise).  It merely has to do with the optical fact that an out of focus point of light will form a circle.

I suggest some folks read more carefully Michael Reichmann's article on this site about "Understanding Depth of Field."
« Last Edit: October 25, 2006, 09:14:44 AM by howiesmith » Logged
jashley
Jr. Member
**
Offline Offline

Posts: 53


« Reply #14 on: October 25, 2006, 09:39:33 AM »
ReplyReply

This article kind of mystifies me.  I haven't seen the "problem" that the author describes.  I regularly make 16x20 prints from 1Ds captures (not Mk II) which exhibit excellent front-to-back sharpness from a few feet to infinity.  This is one of the big reasons I went digital--the potential for greater enlargeability (with acceptable sharpness) compared to film.
Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #15 on: October 25, 2006, 09:48:46 AM »
ReplyReply

Quote
Perhaps you fail to see a relationship because none exists. Circle of confusion is an optical term with nothing to with cameras (digital or otherwise). It merely has to do with the optical fact that an out of focus point of light will form a circle.

I suggest some folks read more carefully Michael Reichmann's article on this site about "Understanding Depth of Field."
[a href=\"index.php?act=findpost&pid=82190\"][{POST_SNAPBACK}][/a]

What is used in the DoF calculations is the CoC Limit (simplistically called the CoC) which has everything to do with cameras. It is a somewhat arbirtary choice for the limit of acceptable sharpness based on conventions. Perhaps you are not aware of the commonly used (and rightly disputed in the article) conventions behind the CoC Limit choice for different film sizes?
« Last Edit: October 25, 2006, 09:52:25 AM by NikosR » Logged

Nikos
howiesmith
Guest
« Reply #16 on: October 25, 2006, 11:02:14 AM »
ReplyReply

Quote
What is used in the DoF calculations is the CoC Limit (simplistically called the CoC) which has everything to do with cameras. It is a somewhat arbirtary choice for the limit of acceptable sharpness based on conventions. Perhaps you are not aware of the commonly used (and rightly disputed in the article) conventions behind the CoC Limit choice for different film sizes?
[a href=\"index.php?act=findpost&pid=82197\"][{POST_SNAPBACK}][/a]

I am quite aware of the optical equations and how to apply them to photography.  "Acceptable sharpness" need not be based on any convention.  It can be personal.  I belive it is a personal thing and don't want it decided for me by Canon, Mamiya, Zeiss, etc..  There is no need to allow anyone else to decide what is sharp for you, unless you don't want to be bothred with that decision.

CoC Limit may have everything to do with a camera, but CoC is a "print" term.  How does a painter make something appear out of focus?  What about slides or computer screens?  Circle of confusion (defined and used well before cameras came along) is an optics term.  Now, the circle of confusion may pass through a camera on its way from scene to screen, but ceratinly has nothing to do with the kind (digital or film) or format (call it sensor size) of the camera.

There are many self described "experts" on DoF, but they refuse to (or can't) answer the question about which term(s) in the optical equations are "format."  Some might even try to tell you optics equations don't work for photography.
Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #17 on: October 25, 2006, 11:20:00 AM »
ReplyReply

Quote
I am quite aware of the optical equations and how to apply them to photography. "Acceptable sharpness" need not be based on any convention. It can be personal. I belive it is a personal thing and don't want it decided for me by Canon, Mamiya, Zeiss, etc.. There is no need to allow anyone else to decide what is sharp for you, unless you don't want to be bothred with that decision.

CoC Limit may have everything to do with a camera, but CoC is a "print" term. How does a painter make something appear out of focus? What about slides or computer screens? Circle of confusion (defined and used well before cameras came along) is an optics term. Now, the circle of confusion may pass through a camera on its way from scene to screen, but ceratinly has nothing to do with the kind (digital or film) or format (call it sensor size) of the camera.

There are many self described "experts" on DoF, but they refuse to (or can't) answer the question about which term(s) in the optical equations are "format." Some might even try to tell you optics equations don't work for photography.
[a href=\"index.php?act=findpost&pid=82210\"][{POST_SNAPBACK}][/a]

   I won't disagree with you. I will just note that subscribing to your point of view does away with any notion of 'calculating' DoF which is fine by me. Perhaps you should convince the article author though  

Not calculating DoF is fine by me. It IS after all a subjective issue. But if you want to calculate it (or approximate it mathematically) you have to put in this silly little parameter of CoC Limit which in film is size dependent amongst other things. You can't have it both ways.Unless you can propose an alternative approach towards calculating it.

I really did not want to start all this DoF discussion, which as I said is quite off-topic but I was dragged in it. It has been beaten to death many times to no avail.

Now its your turn. Do you agree or disagree with my original post about sensor size not being enlarged in any way to produce the final output, thus sensor size is in this respect not directly influencing 'enlargeability'. This was my original assertion and my main disagreement with the article in reference. (I hope you have read both).  I just hate it when discussions get side-tracked.

(BTW, there are other things that bother me in the referenced article. Like the fact that the author is using 100% crops to 'prove' that the engraved DoF zone is not good enough to provide resuts indistinguishable from when the subject is properly focused. This is quite silly IMO since he only 'proves' the obvious. He would get the same results just by examining a piece of film under a loupe, or printing a print large enough to test the sensor used. The author ignores what should be obvious. That DOF scales, even when applied to the format they were calculated for, provide for 'acceptable' sharpness for a specific defined output size. 8x10 usually viewed from a specific distance. Nothing more than that. So he proves nothing we didn't know already).
« Last Edit: October 25, 2006, 11:34:02 AM by NikosR » Logged

Nikos
howiesmith
Guest
« Reply #18 on: October 25, 2006, 11:56:52 AM »
ReplyReply

Quote
I won't disagree with you. I will just note that subscribing to your point of view does away with any notion of 'calculating' DoF which is fine by me. Perhaps you should convince the article author though.

Now its your turn. Do you agree or disagree with my original post about sensor size not being enlarged in any way to produce the final output, thus sensor size is in this respect not directly influencing 'enlargeability'. This was my original assertion and my main disagreement with the article in reference. (I hope you have read both).  I just hate it when discussions get side-tracked.

[a href=\"index.php?act=findpost&pid=82216\"][{POST_SNAPBACK}][/a]

I don't know where you got the idea I was saying one cannot calcualte DoF.  I do it frequently when designing a print.  I was just saying the print designer does not need to have his lens/camera maker tell him what convention he should be using.  Sure, I use the convention that the average person can detect about 1 minute of angle.  I am not "stuck" with that though.  When I make prints for my eagle-eyed son, I close that down a bit.

Yes, I agree completely sensor size has nothing to do with calculating DoF.  I have yet to be required to use camera format to determine DoF.

I will not post any further comments so as to do my part to end this DoF discussion.
Logged
EricV
Full Member
***
Offline Offline

Posts: 132


« Reply #19 on: October 25, 2006, 12:12:57 PM »
ReplyReply

Nikos, DoF calculations for a digital sensor are not particularly different from DoF calculations for film.  In both cases, as Howie likes to emphasize, DoF is a property of the optics and not the sensor (provided the sensor has enough pixels or fine enough grain to resolve the optical circle of confusion).  As you guessed, the pixel pitch provides the connection between pixels and distance for the digital sensor.

If you want equations:

CoC (print) = CoC (sensor) x Magnification

CoC (print) is the blur size on the final print.  The largest acceptable CoC is determined by factors like print size, viewing distance, and visual acuity.
CoC (sensor) is the blur size on the sensor.  This is determined by optical factors like distance to subject, lens focal length, and lens aperture of f/stop.

Magnification is the degree of enlargement needed to go from sensor to print.  What does magnification mean for digital sensors?  Let's assume one pixel on the sensor is printed as one dot on the print.  (If the digital file is resized upwards or downwards before printing, you can include this as an extra magnification factor, but I will ignore it here.)  Magnification is then the ratio of the print size to the sensor size, which is also the ratio of the size of a dot on the print to the size of a pixel on the sensor.  An example will probably make this clearer than words.

Example:
     8 inch print at 300 dots/inch from
     2400 pixels on sensor with 8 micron pixel pitch

Here are two equivalent ways to calculate magnification:

Magnification = print size / sensor size
     = (8 inch) / (2400 x 8 micron) ~ (200mm) / (19.2mm) ~ 10

Magnification = print dot size / sensor pixel size
     = (1/300 inch) / (8 micron) ~ (85 micron) / (8 micron) ~ 10

Why does the physical size of the pixel matter, not just the pixel count of the digital file?  It comes back to the point that the CoC on the sensor is a physical quantity, with actual dimensions in microns not pixels.
Logged
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad