Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Article on landscape focusing technique (including tilt)  (Read 6132 times)
torger
Sr. Member
****
Offline Offline

Posts: 1590


« on: July 26, 2012, 04:34:05 AM »
ReplyReply

I have not found so much about focusing workflows out there designed for practical landscape photography, especially not with tilt involved. Either there are very simplistic descriptions or very complicated (from traditional large format photography). I have tried to make something inbetween:

http://www.ludd.ltu.se/~torger/photography/focus-landscape.html

I hope someone finds it useful.

It also includes a DoF table which I have designed especially for practical use in the field when focusing landscape scenes, both with and without tilt.
« Last Edit: July 26, 2012, 11:14:00 AM by torger » Logged
MichaelEzra
Sr. Member
****
Offline Offline

Posts: 659



WWW
« Reply #1 on: July 27, 2012, 07:51:24 AM »
ReplyReply

Anders, thank you for sharing this article. I browsed through your website - did not know yo were working on audio - very interesting articles there as well.
Do you think any of the audio processing techniques might apply in RawTherapee?
Logged

theguywitha645d
Sr. Member
****
Offline Offline

Posts: 970


« Reply #2 on: July 27, 2012, 12:10:07 PM »
ReplyReply

Interesting article. Some practical advice and thoughts, but many common errors as well.

Actually, diffraction and CoC is determined by the format size. Pixel size has nothing to do with it. Basically, you have not discussed viewing conditions nor even factored them in which makes discussing of sharpness and DoF irrelevant. Sharpness and DoF are perceptual properties of an image, not absolute qualities of the optics.

Also, you DoF diagrams for Scheimpflug should be using curved lines as the increase in DoF with object distance is not linear.

I am really glad to see the images of you working with the Techno and having fun in the field.
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #3 on: August 02, 2012, 04:25:18 AM »
ReplyReply

Interesting article. Some practical advice and thoughts, but many common errors as well.

Actually, diffraction and CoC is determined by the format size. Pixel size has nothing to do with it. Basically, you have not discussed viewing conditions nor even factored them in which makes discussing of sharpness and DoF irrelevant. Sharpness and DoF are perceptual properties of an image, not absolute qualities of the optics.

Also, you DoF diagrams for Scheimpflug should be using curved lines as the increase in DoF with object distance is not linear.

I am really glad to see the images of you working with the Techno and having fun in the field.

No there are no such errors, I'm quite sure I know what I'm talking about :-). You obviously think viewing conditions is central for DoF, which is ok and it certainly is in the end, but I don't think it is useful model for a workflow.

I don't think about viewing conditions when I take a picture, I instead think about maximizing what the camera system can do. I may think about viewing conditions when I choose which camera to use, that is I don't necessarily pick up a high resolution camera if I intend to make a web picture or a postcard print. I think it should be obvious though that the article is about maximizing sharpness of the given system, i e if you have a 60 megapixel camera the goal is to get 60 sharp megapixels, and why you chose a 60 megapixel camera would be because you want to (be able to) print large and/or be flexible with cropping.

Sure one could make tables for intended viewing conditions and end up shooting at f/45 for smaller prints, but then using tech cameras or tilt lenses become rather meaningless. The most popular way to work seems to be to first pick a camera that can deliver the resolution you need for you prints, and then through shooting technique and proper DoF models maximize that resolution, and that is what I aim to provide.

I probably should add a more elaborate discussion about viewing conditions though since you are not the only one to comment on this.

Curved lines for the Scheimpflug DoF wedge? Huh? Are you sure you have not looked at diagrams that have a logarithmic distance scale? Mine are linear. I have written software myself using the established formulas available in books and on the web, I surely would have noticed a non-linear factor :-)
« Last Edit: August 02, 2012, 05:44:41 AM by torger » Logged
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #4 on: August 02, 2012, 04:36:14 AM »
ReplyReply

Anders, thank you for sharing this article. I browsed through your website - did not know yo were working on audio - very interesting articles there as well.
Do you think any of the audio processing techniques might apply in RawTherapee?

I was working on audio ten years ago, I maintain some of the software still though and it is used mainly in the academic world and by some HiFi hobbyists. I was doing a lot of convolution and RT does deconvolution :-). I have use of the skills when trying to penetrate image processing algorithms, but the code is much different. I'd like to contribute more to the RT project than I do, but I no longer have the same work capacity as I had when I was younger. I do lots of algorithm analysis, thinking and coding at my daytime job and that fills the quota.
Logged
theguywitha645d
Sr. Member
****
Offline Offline

Posts: 970


« Reply #5 on: August 02, 2012, 05:56:45 PM »
ReplyReply

No there are no such errors, I'm quite sure I know what I'm talking about :-). You obviously think viewing conditions is central for DoF, which is ok and it certainly is in the end, but I don't think it is useful model for a workflow.

What you need to do then is explain why the science behind this no longer works and your model is accurate. I see nothing to support any change in these basic concepts. And I am quite sure I know what I am talking about. Wink

Quote
I don't think about viewing conditions when I take a picture, I instead think about maximizing what the camera system can do.

That is irrelevant as viewing conditions are relative and always implied in the process. There is no absolute frame where a camera can work outside the idea of a viewer.

Quote
I may think about viewing conditions when I choose which camera to use, that is I don't necessarily pick up a high resolution camera if I intend to make a web picture or a postcard print. I think it should be obvious though that the article is about maximizing sharpness of the given system, i e if you have a 60 megapixel camera the goal is to get 60 sharp megapixels, and why you chose a 60 megapixel camera would be because you want to (be able to) print large and/or be flexible with cropping.
Quote

But sharpness is a subjective quality based on an observer--that is what CoC are all about. To maximize it based on an abstract concept or absolute frame makes no sense. And if you crop, then the CoC change, just as it changes with format. It has nothing to do with maximizing sharpness at pixel level.

The other weakness in your argument is that only the focal plane counts in terms of resolution. However, having insufficient DoF, with or without movements, results in not having enough sharpness and resolution in front of or behind the object.

Quote
Sure one could make tables for intended viewing conditions and end up shooting at f/45 for smaller prints, but then using tech cameras or tilt lenses become rather meaningless. The most popular way to work seems to be to first pick a camera that can deliver the resolution you need for you prints, and then through shooting technique and proper DoF models maximize that resolution, and that is what I aim to provide.

Ironically, all DoF models assume an observer. So you recognize the DoF models are useful, but you do not think viewing conditions relevant. DoF cannot be calculate without a specified viewing condition.

Quote
I probably should add a more elaborate discussion about viewing conditions though since you are not the only one to comment on this.

I would say that is a good idea. Unfortunately, you have to rewrite your entire hypothesis.

Quote
Curved lines for the Scheimpflug DoF wedge? Huh? Are you sure you have not looked at diagrams that have a logarithmic distance scale? Mine are linear. I have written software myself using the established formulas available in books and on the web, I surely would have noticed a non-linear factor :-)

Yes, my mistake.
Logged
buckshot
Full Member
***
Offline Offline

Posts: 187


« Reply #6 on: August 02, 2012, 08:25:37 PM »
ReplyReply

Very nice article Anders, thanks. I wish my English was that good (and it's the only language I speak!)

When it comes to Tilt, I've always found the following short videos helpful (from the great northlight-images website):
« Last Edit: August 02, 2012, 08:30:09 PM by buckshot » Logged
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #7 on: August 03, 2012, 10:30:34 AM »
ReplyReply

There actually is an implied viewing condition, namely that on screen at 100% the CoC should be small enough (but on smaller) so that after sharpening the edge of the DoF is virtually indistinguishable from the PoF. I am not alone using 2x pixel size or airy disk based CoC. The reason for choosing this is that this represents the reasonable peak performance of the system. Since landscape photography is generally somewhat diffraction limited even at optimal aperture (since one wants good corner performance) I think an airy disk based CoC is most elegant.

The thing is that DoF tables are used in the field to make proper  decisions about aperture and tilt etc - the purpose is to answer what settings gives us the sharpest image the camera can deliver with a deep DoF. Not a "sharp enough image for some yet undecided print size".

Sure you can use a DoF model when evaluating actual prints, for example a sports image with PoF  slightly off target will  look in focus if printed small enough or watched far away, but that is an entirely different use.
« Last Edit: August 03, 2012, 11:11:00 AM by torger » Logged
theguywitha645d
Sr. Member
****
Offline Offline

Posts: 970


« Reply #8 on: August 03, 2012, 11:16:35 AM »
ReplyReply

100% monitor view is not a real world viewing condition. "Peak performance of a system," as you call it, is irrelevant and well as false--it is only a "peak" in terms of resolution, but not DoF, for example--you are just picking one subjective criteria over another. What you need is the best solution for a pleasing image--resolving more than a viewer can see is rather pointless, especially if you don't have the "best" DoF to make the most pleasing image.

But your article is simply personal opinion. You are presenting it is some sort of undisputed fact and the people who actually understand DoF and CoC are simply wrong.

(You are also repeating another myth that print size is an issue. If you really understand the concepts of viewing distance you would know that size is not important, which is why print size is not a variable in DoF calculations.)

The perception of an image is really important because we look at images. So, to state that 100% monitor viewing is the valid way to determine sharpness is simply false because the observer is not going to do that. It also ignores the robustness of the illusion--photography is an illusion.
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #9 on: August 03, 2012, 02:31:58 PM »
ReplyReply

I'll try to explain why the "old science" is no longer adequate - it is basically about that people are no longer shooting on film, and with digital the expectations have changed. A 24x36mm D800E sensor can deliver prints which in terms of grain-free resolution is very similar to 4x5" film.

Sharpness has always been subjective. Even in the film days many thought that the typical CoCs were too large, because it was based on the assumption that people did not have very good vision or did not really step close to a fine art print. The traditional film CoCs were 0.1mm for 4x5" and 0.03 for 24x36mm, this corresponds to a resolution of perhaps 4 megapixels or so at the DoF edge (!).

It is very obvious that if you take your D800E 36 megapixel camera and base hyperfocal focusing on a 0.03mm CoC the skyline in the horizon will not be critically sharp when you print big and step close.

To be loud and clear - DoF calculators/tables which uses traditional film CoCs are next to useless in the field if you are into high resolution "full DoF" landscape photography. Instead of continuing to explain why it is so useless, I'll try to explain why my method (which certainly is not unique, many use small CoCs these digital days) is good:

My DoF tables aims at a basic goal - that all pixels in the whole scene should be rendered equally sharp, that is when you nose the (sharpened) large print it should be virtually impossible to distinguish the plane of focus from the edge of the DoF. This leads to much narrower DoFs than traditional, but still very much practical you can fulfill this goal without too much diffraction tradeoff even with 40-60 megapixel systems. It is practical and well-defined, anyone understands what the DoF edge means, it means the edge where one begins to see a difference from PoF at maximum enlargement.

Some use 2x pixel size as CoC but actually I think the CoC = Airy disk is more elegant, resolution independent and therefore future proof. That concept is based on the assumption that the optimal lens aperture is somewhat diffraction limited, which it is for deep DoF landscape photography since you want good corner-to-corner performance. I don't see anything that indicates that this balance between lens and sensor resolution will change in the near term.
When aperture is reduced farther the diffraction increases and the CoC should increase too since the edge where one starts to see a difference moves. The actual edge if studied critically is probably somewhere around 0.8 of the Airy disk diameter, but I think that is hair splitting.

Another way to describe the same thing is that the DoF is balanced so that there is no further sharpness gain at the edge if you reduce the CoC. This is a very important property - the CoC is perfectly balanced to achieve peak performance for the DoF you need. You don't stop down more than you need to get everything equally sharp, and the CoC is small enough to avoid the nasty surprise that what you thought should be pixel-sharp really is not.

We assume here that the photographer is aware of how much diffraction affects her/his system, and I do think that is true for most landscape photographers with some experience. The photographer makes a subjective decision of how much f/22 hurts for example to gain DoF, and then it makes perfectly sense that the CoC should follow and match that. One could make all sorts of subjective fine-tunings like having 0.8xAiry disk for larger apertures and 1.5xAiry disk for f/22 since one have already lowered the expectation then, and everyone is free to do that. I do it by estimation by at very small apertures let some things be a bit outside the DoF if necessary.

Holding tight to the old CoC and viewing models all this risk to end up with a Ken Rockwell-esque "4 megapixels is all you need for any use" type of statement, because that is what the typical result of viewing distance calculations is (which is what the old CoC is based upon). But even with my limited experience of fine art printing and display, and high quality photo book prints, I know that in practical viewing in homes and at galleries you can appriciate much higher resolutions than those traditional models lead to. Far from all people nose the prints or even care about image quality in general, but some do, the photographer herself/himself not the least.

You say I don't understand the viewing distance concept, I understand it very well. The traditional thinking which do lead up to "4 megapixels is all you need" is flawed. Make a normal size print of a flower and then make a large size print of a detailed landscape, perhaps a city at a distance. Hang them on the wall and look how people that look at the prints behave. For the flower they will behave nicely and stand at a distance so they can take in the whole image at once. For the detailed landscape they won't be so nice. They will step up close and inspect those interesting small details (if there's something there to find). 100% pixel peep is quite real to represent what will be visible for this type of print done at 150-200 ppi. For high quality photo books the same things happens but in a smaller format. The viewing condition is very much relative depending on subject, print size and presentation. Detailed landscape scenes printed big presented in a gallery mounted at comfortable viewing height is an example where the prints will face very close inspection indeed.

I have a 33 megapixel system. When I shoot I generally don't know where the picture will end up, but I preferably want to come home with files that can be printed as sharp as possible in large sizes. Otherwise I'd use a much lower cost and lighter hand-held system with less megapixels (which I do for other types of photography). Tables based on the DoF/CoC model described above works and helps me achieve this goal, and I don't think I'm unique in having this goal, but rather I think it is what landscape photographers with high resolution gear generally want -- that is getting the sharpest possible image the gear can provide.

Obviously there are situations when one has a more relaxed relationship to the DoF, for example if we don't hyperfocal or tilt but actually put the focal plane at the main subject and it clearly is the main one then we don't need to worry as much of how sharp the DoF edges are. In any case the traditional CoCs are not delivering what I and many others expect, so we need something other, and I think I have (and the others with similar solutions) have come up with a good way that actually works is easy to understand and gives expected results, and is also a much better basis to use for "relaxed estimations" than CoCs that are heavily relaxed in the first place.

Ok this was long, but I believe very much in this model and I'm using it with good results, and I see others use similar models.

The thing that is always asked is "how to make the image as sharp as possible?". Use a stable tripod, use mirror up (if applicable), cable release, use high quality lenses, use proper focusing technique. My article is about focusing technique. Using old-style CoC for deep DoF is not adequate for that purpose. The question is not "how to make the image as sharp as needed?", which the traditional CoC idea tries to answer (focusing technique will be exactly the same though, only the values in the tables will change).
« Last Edit: August 04, 2012, 03:20:47 AM by torger » Logged
Les Sparks
Full Member
***
Offline Offline

Posts: 213


WWW
« Reply #10 on: August 05, 2012, 07:44:33 PM »
ReplyReply

Thank you for the article.
Les
Logged

erpman
Jr. Member
**
Offline Offline

Posts: 61


« Reply #11 on: September 15, 2012, 04:20:07 AM »
ReplyReply

Hi Torger

Great article, I read it with great interest and totally see where youre coming from in your proposal to redefine the expectations for DoF in the digital era. I struggle with these issues all the time, as my work involves huge detailed prints, that people indeed walk up closely to. I simply do not want unsharpness in my images, I think its "un-photographic".

Certainly, people are calling you out on your strict demands for sharpness at 100% screen viewing, but once you enlarge an image 2-300% this is no longer so strict anymore.

I especially took interest to your dynamic approach to setting the CoC relative to aperture, but had some difficulty understanding the practical implementation of the two techniques. Ill give two examples to see if I got it right, if not, please correct me  Smiley

Tecnique 1, CoC 2x pixel size: If I use a d800E the pixel size is 5m, and the CoC will thus be 10m (!) This seems unpractical as you may end up stopping down way too far.

Tecnique 2, dynamic approach based on airy disk size: at f11 the airy disk is 14.7m, and the CoC for f11 will be 15m. In practical application, this means going back and forth a little? This means that if you start with a given aperture and find that it will not give you sufficient DoF at its given CoC you stop down, and at the same time increase the CoC to reflect the new aperture??? Some examples of this particular techique would certainly be useful!

Thanks!
« Last Edit: September 15, 2012, 04:38:41 AM by erpman » Logged
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #12 on: September 20, 2012, 02:08:17 AM »
ReplyReply

Tecnique 1, CoC 2x pixel size: If I use a d800E the pixel size is 5m, and the CoC will thus be 10m (!) This seems unpractical as you may end up stopping down way too far.

Tecnique 2, dynamic approach based on airy disk size: at f11 the airy disk is 14.7m, and the CoC for f11 will be 15m. In practical application, this means going back and forth a little? This means that if you start with a given aperture and find that it will not give you sufficient DoF at its given CoC you stop down, and at the same time increase the CoC to reflect the new aperture??? Some examples of this particular techique would certainly be useful!

Your examples are correct. The appendix now has a picture showing how the dof edges looks with CoC = airy disk. The edges are indeed slightly softer, but very little so, hardly visible in a print even with loupe after sharpening. If you do not want any softening at all you need to drop down to about 0.5 x airy disk diameter, but that becomes very hard to work with.

With technique 2 you use pre-calculated tables so you don't need to fiddle around with the CoC. It will be a mess to use if you have a DoF application where you need to change CoC in it each time you change the aperture of course. The idea of technique 2 is that there's no value in having a CoC that outperforms diffraction, since diffraction will mask that anyway -- so the more diffraction you got the larger CoC you use.
Logged
erpman
Jr. Member
**
Offline Offline

Posts: 61


« Reply #13 on: September 20, 2012, 03:43:55 AM »
ReplyReply

I thought so, in my Iphone dof-application you can indeed change the CoC quite quickly, but it seems better to have it ready at hand. The coolest thing would be if there was an DoF application where you could set the CoC specifically for each aperture, but I doubt that it exists.

Would you care to give an example of how such a pre-calculated table might look like?
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad