Luminous Landscape Forum
Equipment & Techniques => Landscape & Nature Photography => Topic started by: Rikard_L on May 15, 2012, 10:26:41 am
-
Hi all,
I plan on trying to make a landscape image by stitching like 3x3 shots together. One thing I have been thinking about is how to make the final result to look sharp all over. If the foreground is near enough the depth of field will be quite shallow even if I shoot at F22. Is there maybe a way to combine this stitching with focus stacking?
Like:
Lower 3 shots: Focus on the foreground
Middle 3 shots: Focus on the middle ground
Upper 3 shots: Infinity focus.
And then stitch them together with some sort of tool, like Hugin.
Altering focus will probably cause some problems because of focus breathing?
Any tips on technique and/or tools I should know about are greatly appreciated.
-
I use focus stacking regularly (Helicon). I also stitch. But so far not combined on one image.
My instinct tells me:
- Make precise focus marks on your lens (tape or whatever) so that your focus increments are 100% repeatable from shot to shot.
- Do all the stacking before the stitching.
Good luck! Share your results. I've been meaning to try this, too.
Rand
-
I use focus stacking regularly (Helicon). I also stitch. But so far not combined on one image.
My instinct tells me:
- Make precise focus marks on your lens (tape or whatever) so that your focus increments are 100% repeatable from shot to shot.
- Do all the stacking before the stitching.
Good luck! Share your results. I've been meaning to try this, too.
Rand
Thanks a lot for sharing your ideas Rand. I'll definitively look into the Helicon software.
My initial though was that I could focus on the foreground and shoot, 3 shots left to right. And then move the camera up to the middle part and refocus, again 3 shots from left to right. Likewise for the top part.
And then in post, while stitching, the overlapping parts could somehow be "focus stacked". But I guess I have to focus stack each sub-composition, before stitching them together?
-
I think you could try it either way, but I think the stitch might end up cleaner if it is working from "already stacked" segments.
This is merely an educated guess on my part, though. That's why I recommend having your focus increments marked / precise.
Keep us posted!!
This is a redundant posting of this image, but I'll put it here to show what I routinely achieve with Helicon Focus.
Shot at 200mm f/8 - five shot stack.
(http://rsadams.smugmug.com/Landscapes/Landscapes/i-h6QPWcw/0/L/stacked-trees-L.jpg)
Rand
-
I agree. Stitch last.
I've done a couple of tests where I did focus stacking, exposure blending and stitching. It's a lot of work, but it can be done. I did the focus stacking in CS5 with "auto blend layers". I manually did the exposure blending with layer masks and stitched with CS5, avoiding like the plague "auto" mode. Cylindrical seems to work best.
The main lessons I learned were
1) "auto blend layers" leaves tiny borders between subjects on the stacked layers. Visible at high magnification, but tolerable. repairable manually
2) you need far more focus layers images than you think
3) I look forward to the days when HDR no longer necessary
4) the DAM issue is critical when you have a dozen or so images destined for a single resulting image.
-
"3) I look forward to the days when HDR no longer necessary"
We're getting close now w/ LR4.x! ;D
-
Apologizes for my ignorance, but what is this all about? It's not better, faster and easier doing a shot in the closest f number than focusing in different planes, then merging them? What's the advantage of doing it by planes? as long as I know, far objects which are portrayed won't hold details, so what's the deal with focusing the far objects if camera sensor isn't able to offer a great focus on those?
Thanks in advance,
Ø3 :)
-
Apologizes for my ignorance, but what is this all about? It's not better, faster and easier doing a shot in the closest f number than focusing in different planes, then merging them? What's the advantage of doing it by planes? as long as I know, far objects which are portrayed won't hold details, so what's the deal with focusing the far objects if camera sensor isn't able to offer a great focus on those?
Thanks in advance,
Ø3 :)
I'm not sure I fully understand your question, but let me take a stab at it. If you look at the image of the trees I posted and realize that it was shot at 200mm, you can easily imagine that even at an f-stop where diffraction becomes a real issue e.g. f/32 I'd never achieve front to back focus at that focal length. So, focus staking allows a "different view" of subjects than is possible otherwise with the typical FF camera. Another advantage is that one may use the "sweet spot" in the lens aperture for maximum sharpness w/o concern for depth of field "front to back."
With wider lenses, e.g. my Zeiss 24, there isn't much, if any advantage to focus stacking once stopped down a bit and using hyperfocal distance focusing. But having the technique available adds a tool to the vision toolbox, IMO.
Is that what you were asking?
-
I'm not sure I fully understand your question, but let me take a stab at it. If you look at the image of the trees I posted and realize that it was shot at 200mm, you can easily imagine that even at an f-stop where diffraction becomes a real issue e.g. f/32 I'd never achieve front to back focus at that focal length. So, focus staking allows a "different view" of subjects than is possible otherwise with the typical FF camera. Another advantage is that one may use the "sweet spot" in the lens aperture for maximum sharpness w/o concern for depth of field "front to back."
With wider lenses, e.g. my Zeiss 24, there isn't much, if any advantage to focus stacking once stopped down a bit and using hyperfocal distance focusing. But having the technique available adds a tool to the vision toolbox, IMO.
Is that what you were asking?
Oh, okay! Now I got the point! I thought the image was taken with another type of lens. thank you very much for the clarification!
Ø3 :)
-
Recent discussion that addresses this topic...
http://www.luminous-landscape.com/forum/index.php?topic=62513.msg504045#msg504045
-
Recent discussion that addresses this topic...
http://www.luminous-landscape.com/forum/index.php?topic=62513.msg504045#msg504045
Excellent! Thanks for linking...
Rand
-
2) you need far more focus layers images than you think
Right on!
And, what surprises a lot of people is that if your shot contains lots of information at or near infinity, you can benefit from having as many as two steps that are just barely short of infinity. Including steps at 50 feet, 200 feet, infinity is not a wasted effort when shooting f8 with medium FL lenses.
It's kind of counter intuitive, but in terms of the amount of lens rotation needed you can have relatively wide steps for nearby objects, but you need increasingly fine steps as you approach infinity.
And while it is possible to make a set of focus marks that will cover a range of focus for objects from nearby to infinity, there are often cases as with the tree shot where scenes naturally organize themselves in distinct planes, in which case just march your focus from important plane to important plane.
Oops, somebody just pointed out my old post which I had hoped would die because of the huge download was murdering my bandwidth! Look if you must, but please don't reply to that article which will top post it! But please realize that's the really hardcore solution. What didn't get sufficiently stressed there was that you need to test finely separated focus lanes near infinity.
PS, don't anybody think they can shoot that tree shot with a T/S lens.
-
One thing I have been thinking about is how to make the final result to look sharp all over. If the foreground is near enough the depth of field will be quite shallow even if I shoot at F22. Is there maybe a way to combine this stitching with focus stacking?
Forget F22, the diffraction at this aperture will makes it unsharp.
Focus stacking at the optimal aperture - around F8 (depending on your lens and sensor size) is a better method.
-
I have tried focus stacking on flowers with different programs and one was Helicon Focus. I can't recall the other, but neither worked very well. They both produced a lot of ghosting. I was told that PS, which I don't have, does a better job than the independent programs. But frankly after what I have seen, I doubt it. It worked best when the distances were minor and not great, like say, 1/4 inch difference between petals. That's been my experience.
JMR
-
I don't do a huge amount of focus stacking but I did compare the two year old Helican software I have with stacking in CS5 and found no difference that I could see.
-
I have tried focus stacking on flowers with different programs and one was Helicon Focus. I can't recall the other, but neither worked very well. They both produced a lot of ghosting.
Hi John,
The ghosting may be related to your choice of Method A or B. Recently a Method C was added. These 'methods' are the strategy that will be used to select the best focused areas from the individual focus brackets. It might be necessary to combine the output from several of these methods (which is part of the retouching functionality in the Pro version) to get the best results.
Photoshop, as usual in these stacking operations, offers no possibilities to intervene, and when I tried it produced (not bad, but) inferior results to the Helicon solution.
Cheers,
Bart
-
Almost every pano I make now uses stacking and I see HUGE differences between the quality of stacking in CS5 and Helicon. While CS5 can stitch almost as good as any specialized program, it's got a ways to go on stacking, which in PS is just a variation on blending.
If you're getting ghosting, you need to shoot more focus planes and play with the parameters. For stacking normal landscapes usually 3 to 7 planes will do the job, but for objects very close you'll need more. Macro and microscope shots may use 100 or more planes.
I have been told that Zerene may be better for extreme closeup work, although I have no experience with that. Helicon does seem to be dramatically faster and gives me predictable, works-every-time results.
-
PS, don't anybody think they can shoot that tree shot with a T/S lens.
Since the trees in my shot stagger left-right-left-right-&-center, I think you'd be hard pressed to do this w/ TS.
Unless there's something I'm missing or a new waffle-Scheimpflug technique I've missed. ;D
-
I have a noob question:
Instead of using stacking software or algorithms, why don't people use masking in photoshop to 'layer in' each plane? Won't that give precise results, or is there a downside?
-
Instead of using stacking software or algorithms, why don't people use masking in photoshop to 'layer in' each plane? Won't that give precise results, or is there a downside?
Photoshop can actually do focus blending through its Auto Blend Layers function. From a set of focus brackets arranged in layers, it attempts to blend-in the parts of the sharpest image in a particular area. But it's not very subtle, and it leaves you with complex sets of masks that need refining.
For landscapes that are mostly flat planes leading out to the horizon, using hand painted mask can work OK. But there will be magnification differences in each image due to focus breathing, and the overlay areas will not be perfect unless you spend some time scaling each layer, which is not an easy task. I did this once in a while many years ago, and when the focus stacking programs started to appear I was overjoyed and started to use stacking much more.
But for complex images with nearby objects like tall bushes etc, it is much easier to use programs like Helicon and Zerene. They automatically scale each of the bracket images to perfectly overlap, and if you get the parameters right there will very minimal ghosting around nearby objects, which in every case will be better than you can get from Photoshop blending. They output finished images that are ready to use with no further editing due to stacking issues.
-
I have a noob question:
Instead of using stacking software or algorithms, why don't people use masking in photoshop to 'layer in' each plane? Won't that give precise results, or is there a downside?
Hi Sareesh,
I agree with Bill. When you focus at different distances, the magnification factor of the projected image on the sensor will change. Therefore the software must do several things, resize each image, align the images in the stack, adjust variations in exposure, and select/mask the sharper portions of each image, and blend the selection between the successive images to a seamless transition.
That's a lot of (tedious) work to do by hand, especially when the number of images in the stack increase. I've done macro images with more than 100 stacked images each. What's more, programs like Helicon Focus use better resampling algorithms (Lanczos) for these small scale image size adjustments than Photoshop (bicubic) does, and they can use things like a dustmap to automatically remove the shadows of sensor dust particles, which will otherwise leave very noticeable trails of spots due to the resizing.
Cheers,
Bart
-
Thank you for explaining, Bill and Bart. I didn't consider the agony of having a 100 images to blend in!
-
I tried Zerene and Helicon, each left ghosting. And yes this was for macro work of flowers, mostly creative work. At micro levels, you have to be discerning about what to include in or out of focus. It's a different proposition for landscapes. You mostly want everything in focus. What I did not try was combining large numbers of images. I used only 5 or 6 at most. I am just going to have use the old method of trying different focal points and different apertures to get what I want. With film slides, you didn't waste film!
JR
-
I tried Zerene and Helicon, each left ghosting. And yes this was for macro work of flowers, mostly creative work. At micro levels, you have to be discerning about what to include in or out of focus.
Hi John,
Maybe therein lies the problem, just missing the required shots for a smoother transition. The usual issue is with occluding features. When focusing on features further away, foreground features become somewhat transparent and the blending has trouble deciding which sharp image to use, foreground or background. HeliconFocus' Method B (using a depth map) is better at getting the edges right, because Method A uses averaging which is better for surface structures. The recently added Method C (pyramid) may also work well. It can work well if you combine parts of the the A/B/C output by using the retouching functionality of HF Pro. Zerene also offers retouching capability if I remember correctly.
It's a different proposition for landscapes. You mostly want everything in focus. What I did not try was combining large numbers of images. I used only 5 or 6 at most.
Depending on the magnification factor, that may be too few. Based on the magnification factor, you can use the (slightly less accurate, but accurate enough) DOF approximation formula for close-up and macro photography (http://en.wikipedia.org/wiki/Depth_of_field#Close-up), which uses the magnification factor as input. Your sensor size in relation to the size of your object will give you the correct magnification factor to use. The CoC to use in that formula equals the sensel pitch, if you want seamless sharpness per pixel from front to back. If you are going to output at less than pixel per pixel size, you can increase the CoC proportionally.
Cheers,
Bart
-
Basically, if you have nearby objects like trees with bare branches where you can see the horizon through openings between the branches, you need to shoot as many focus planes as possible to keep the ghosting down.
However, ghosting is not always bad. Sometimes very subtle ghosting around nearby vegetation can help separate it from more distant textured background in ways that is really not objectionable in the print.
-
I have a noob question:
Instead of using stacking software or algorithms, why don't people use masking in photoshop to 'layer in' each plane? Won't that give precise results, or is there a downside?
Using Auto-Align will match up your layers so there are no offsets.
There are many photographers on the Nature Photographers Network - http://www.naturephotographers.net/enter.html
that are churning out superb results using combinations of stacking, stitching, and exposure blending.
NPN member Marc Adamus comes to mind - http://www.marcadamus.com/ .
NPN member Tony Kuyper has produced some excellent actions for creating Luminosity Masks - http://goodlight.us/ , which are quite beneficial. Check out his tutorials.
NPN member Pat Sampson, who is "taking a break" has some exquisite florals that are stacked - http://www.naturephotographers.net/imagecritique/ic.cgi?a=up&pi=PATSAMPSON&ns=1&CGISESSID=bb8866b4fac309c6645e8ea0aa50bf59&u=24287
My favorite Pat Sampson image is titled "Red Tulips" five rows down.
Here's one by NPN member Craig Strand I really like - http://www.naturephotographers.net/imagecritique/ic.cgi?a=vp&pr=179350&CGISESSID=bb8866b4fac309c6645e8ea0aa50bf59&u=23948
Here is a link to a few more NPN stacked florals - http://www.photoportfolios.net/portfolio/pf.cgi?a=sp1&gr=4
Note - select "Flowering Plants," then type "image stack" in Optional Keywords.
Developing good skills through practice can produce amazing results.
-
Thanks for the links, Michael. Will check them out.
-
Thanks Michael, I hadn't looked at that site for years. Great work.
-
Great thread for us noobs: thanks, everyone. This explains some of my soft landscapes and I look forward to trying these techniques.
Scott
-
Hi folks,
Just to let you know, I've started a thread about a new web-based tool (http://www.luminous-landscape.com/forum/index.php?topic=81485.0) that I created. The 3rd section of the tool also assists in calculating the required number of slices, and their focus distance, for the creation of a perfectly focused stack. The results can also be a bit sobering for those who want everything in perfect focus, but you can always reduce your output size ... ;)
Cheers,
Bart
-
When considering combining focus stacking and exposure blending for Canon SLR camera there is another approach to consider which can more or less automate your work in the field. I am talking about dslr Controller, an APP that runs on later versions of the Android operating system, meaning your phone and tablet. It is in beta ($9) but it appears to work very well as is. This APP is not a one trick pony. It can bull focus from point A to B, automate and extend HDR capture in many exposure steps, do both at the same time (lots images), it will calculate the number of exposure steps needed for focus stacking with certain restrictions. With a tablet and a USB extension it will remote most of the functions of your Canon SLR to the display on the tablet. You can use various touch commands on the tablet to change f stop, ISO, etc. See http://dslrcontroller.com/ for details. Magic lantern can do similar things (plus more) on certain Canon dslr's and it is free. Pros and Cons. I tried both and thought dslr controller was better for the focus stacking kind of apps.
-
I tried both and thought dslr controller was better for the focus stacking kind of apps.
Hi,
Indeed, Dslr Controller gets good reviews from users. Another alternative is Helicon Remote (http://www.heliconsoft.com/heliconsoft-products/helicon-remote/) for Android (also available for Windows and Mac). It can also interface with e.g. the motorized StackShot focus rail.
Cheers,
Bart
-
I tried Zerene and Helicon, each left ghosting. And yes this was for macro work of flowers, mostly creative work. ... What I did not try was combining large numbers of images. I used only 5 or 6 at most.
For close-up work (not macro) I'd need more like 30-50 images.
Back in June I made series of close-up flower images and was very disappointed with the ghosting around each petal in the merged image. It was a frustrating couple of weeks, of blindly working through different apertures in the hope that might have something to do with the problem. I must have had hay-fever.
Once I'd exhausted all the possibilities I finally re-read the Enfuse documentation and found the very clear instruction that if the merged image showed visible seams then use different contrast-edge-scale and contrast-min-curvature parameters. Doh.
And lo, ghosting was reduced to a scary story, and there was much rejoicing.