Pano's, should I be refocusing?

Hi,

I've now completely replaced my LF gear for a pano setup. I'm head over heels in love with the 6X12 format and find that I can achieve that easily with a 3 shot pano using my 5D. That gives me a native 18 megapixel file. Great. Today I tried shooting it in vertical and just shooting more frames to achieve the same ratio but with more more resolution. With a 6 shot pano I've hit 39 megapixels and the resolution is just great, more than I need technically but it's so little effort, why not?

Apart from the fact that when stitching like this you get a lot of wide angle effect due to being so much nearer to the subject (I would love to use longer lenses but DOF is a serious problem), evenso with my 50mm, the difference between the focal point used for the 'middle' frame and the subject when you are rotating to the far left and right is quite dramatic. I'm shooting at f22 anyway but diffraction is rather obvious this stopped down and to be honest with that big a shift in focus, I'd prefer to get it right rather than rely on DOF to 'see me through'.

As such, should I refocus on the subject each time I rotate making sure that each frame has the optimum focus on the subject or will I start getting problems?

I do know about programs such as Helicon but trying to avoid the need (it's useless for moving subjects such as branches anyway, autopano pro is actually very good with that), I would prefer to make the most of each frame focus wise in camera and leave helicon for territory needing a tilt lens!

Ben Rubinstein said:

As such, should I refocus on the subject each time I rotate making sure that each frame has the optimum focus on the subject or will I start getting problems?

Click to expand...

Hi Ben

yep, you will get problems when stitching two different focuses, in the seamed/transition aerea of the two images/focuses. Personally, I never looked at it to be a problem, if in a photographic image whatever, some things were out of focus.

Some hints:

- reshoot a really important frame - after the usual stitchin shots - but with a slightly different focus. °Combine° that frame prior to stitching into the same frame, with the °stitch focus°, in PS, with layermasks. Now use the combined frame for stitching.

- F 22 is creating diffraction, with the 5 D, AFAIK 11 - 16 is the limit: so use better a 35mm or 28 (on FF) instead of the 50 mm.

- Shooting the pano twice, with different focuses, stitching them individually, and combining them in PS later; layermasks, again.

The distagon 28 & 35 are my favourite lenses for stitching, just because of your problem, FF too.

What are you stitching?
Landscapes?.... then you might try the cylindrical projection.

I'm not understanding: "Apart from the fact that when stitching like this you get a lot of wide angle effect due to being so much nearer to the subject"

Any image examples?

Ben Rubinstein said:

As such, should I refocus on the subject each time I rotate making sure that each frame has the optimum focus on the subject or will I start getting problems?

Click to expand...

With some careful planning, one can re-focus between images. The important thing is to get the DOF between images to connect. This may require a very liberal overlap between tiles, depending on angle. It also requires to 'disconnect' the focal length settings between images in the stitcher, because re-focusing will change magnification which is in turn controlled by the focal length parameter in the stitcher. I can also help to use landscape orientation for the tiles of 'floors/ceilings', and portait orientation for 'walls'. Here is an example of a refocused, stitched, landscape (not by me): http://www.clarkvision.com/photoinfo/large_mosaics/index.html.

Bart

Hi,

I'm shooting urban landscapes, let me make it clear that everything in the image is already within the DOF of the chosen aperture, there won't be any OOF areas, my idea to refocus on the subject is just to keep the area of ultimate sharpness on the subject rather than letting it remain sharp through DOF alone.

Here are two photos, the first shot with a 3 frame stitch, the 2nd with a 6 frame vertical camera orientation. The lens was the same, my 50mm macro, designed for stop down performance. Please pardon the hugely mundane subject and varying time of the day. As you can see, eventhough the vanishing point isn't done as well on the 2nd picture, being closer to the subject introduces the higher 'WA' distortion. Not really a problem in real world shooting, I'm going for asthetic not accurate urban landscapes, the overhang really makes this look worse than it would be usually.


I think I'm going to have a try later today, once with refocus, once without, see if there are any issues with that much DOF. The 2nd image, the 39 megapixel one, wow but that thing has resolution when properly sharpened, blows your mind away and a 6 frame stitch is so quick to do!

And yes, I really have to get round to registering my Autopano Pro, been leaving it till I really got a hang on stitching, the images aren't going anywhere... ;-)

Ben Rubinstein said:

I'm shooting urban landscapes, let me make it clear that everything in the image is already within the DOF of the chosen aperture, there won't be any OOF areas, my idea to refocus on the subject is just to keep the area of ultimate sharpness on the subject rather than letting it remain sharp through DOF alone.

Click to expand...

Okay, then there should be no real issue with refocussing. Just be aware that a too narrow aperture will reduce overall sharpness due to diffraction, even for the ultimate sharpness in the actual focus plane.

The 2nd image, the 39 megapixel one, wow but that thing has resolution when properly sharpened, blows your mind away and a 6 frame stitch is so quick to do!

Click to expand...

Yes, having access to all that real resolution gives the subject increased realism. The material/surface structures almost become tactile. Adding focus layer blending techniques (Helicon focus, TuFuse Pro) to pano stitching allows to even beat the laws of physics.

Bart

Ben

You are entering a whole new world! I'm sure that you will enjoy. With a bit of care, even hand held panos can be stitched erally well and can give very easily huge resolution. The original files of my couple of Scottish panos (see here - http://www.openphotographyforums.com/forums/showthread.php?t=6098) are both well over 40Mp, and 40Mp in the shape I want!

I've recently started a little project to make prints at 600dpi without interpolation (except Bayer of course!). That is the camera files sent to the Canon printer at 600dpi gives about 7.5 by 5 from a 5D and 9 by 6 from a 1Ds3. The out put (depending on the input) is a little gem, and I'm hoping to shoot a few panos in Norway that will be suffiently high resolution to extend this up to 16 inches on the short side.


Bart

I've read positive comments on Tufuse. Could you comment on what it can do and how well it achieves this. Also on how easy it is to use and where to get a copy.

Thanks

Mike

Mike Shimwell said:

Bart

I've read positive comments on Tufuse. Could you comment on what it can do and how well it achieves this. Also on how easy it is to use and where to get a copy.

Click to expand...

Hi Mike,

TuFuse is similar to Enfuse, in that they are both free command-line applications that use the same underlying principles of multiresolution blending. There are several GUI applications (for Mac and/or Windows) that allow easier access to the built-in functionality.

TuFuse is different in the sense that it was made with panorama stitching in mind! It automatically does 2 passes over the panorama tiles, one to get exposure blended results in the overlapping regions, and another to get focus blended results with the same image tiles (exact functionality is subject to user selectable parameters). It assumes, just like Enfuse does, that the tiles are already aligned and in optimal registration. TuFuse can e.g. be called from the PTAssembler stitcher (for Windows only) which can take care of the prior alignment.

An enhanced standalone GUI version called TuFuse Pro is to be formally released soon, but the beta version is already fully functional (watermarks are removed after registration, which transfers to the official release). The latest beta 0.95 has more or less fixed a clipping issue that could occur with focus blending, and seems to run very well even on my memory starved PC. That suggests that it efficiently uses memory, even on stacks of 21MP EOS-1Ds Mark III files.

I can recommend TuFuse Pro, as a critical happy user, it makes things so much easier (previews and all). The only thing really missing from the program is auto-alignment, but I already have PTAssembler for that (and can use the TIFF images it outputs), or one can use Photoshop (or the Hugin automatic stack alignment program) to do the pre-alignment. I have no connection to the programmer (photographer Max Lyons) other than having purchased both PTAssembler and TuFuse Pro, and the prices are modest IMHO.

Bart

P.S. Ben's original question doesn't necessarily require a focus blending program like TuFuse (Pro), since 'Smart blending' can already achieve some of that. Things would become easier to improve when using a dedicated focus blender with limited DOF images.

Hi Bart, I do realise that stopping down to f22 will cause diffraction, it's the reason I bought the Macro, it's the least diffraction limited lens of that focal length in the lineup being optically designed to stop down. Diffraction is of course sensor based but there are certain lenses which are designed with that in mind, all the goodness is put in the far end rather than the usual f8 of most lenses and macro lenses fall into that catagory. It's certainly noticeably better at f22 than my 50mm 1.4.

I do realise of course that being able to open up will better the resolution but a lot of my subjects (close up architecture) need a lot of DOF plus have foliage etc in which counts out both focus bracketing and wider apertures. To be honest I've rarely found f22 to be a limiting factor in real world use on a FF camera, f16 is really the limit of diffractionless territory on FF and this is only just one stop worse. Seems to be a case of theory outweighing real world use even if it is noticeable in tests. The real problems I've seen have all been lens based with most primes (and zooms for that matter) just not optimised for small apertures.
Certainly 'DOF sharpness' is going to be less than 'plane of focus sharpness' even if the aperture is less. To put it into perspective we are talking about compositions that I used to need f64 as a minimum on 4X5.

The best thing about all this is that a 6 frame stitch is so fast to do that you can do it at varying apertures and just choose the sharpest that holds sufficient DOF when you get home. I've done this in the past with UWA where it's often hard to imagine in the field that you will have enough DOF at f8 but when you get back you find that your f8 frame was the sharpest and had back to front focus. The picture below:


is an example, shot at 17mm f8 and noticeably sharper at that aperture (lens based) than the f16/f22 I did it at and pixel peepingly sharper even than the f11. Have to be honest, the light was perfect, the air movement non existent otherwise I wouldn't have bothered. Used to use this method (bracketing f/stops to maximise sharpness while retaining DOF) sometimes even with medium format and my favorite 210mm lens with landscape. This second photo is an example, shot on 645 with a 210mm lens and the nearer branches were pretty close. I shot it at f22 and f32 and according to the math of DOF it is actually impossible that it is all razor sharp and in focus back to front at f22 but it is, I didn't know that at the time though.


Mike, can I ask why you are bothering with 600dpi? I thought the eye can't see more than 400dpi anyway?

Ben Rubinstein said:

Mike, can I ask why you are bothering with 600dpi? I thought the eye can't see more than 400dpi anyway?

Click to expand...

Most Canon printers print at a native 600 PPI resolution. Intermediate ink colors are produced by positioning/mixing the smaller ink droplets at higher spatial densities (DPI). Other PPI inputs get resampled by the printer driver to 600 PPI on most paper settings. The difference in resolution is, depending on subject matter and paper choice, visible when compared to lower PPI input. In addition, it allows to apply print sharpening at 600 PPI which produces sharper results.

Visual acuity is, on average, limited to something like 300-400 PPI, but there is also something like vernier acuity which allows to see smaller detail than can be readily resolved.

Bart

Bart, you took the words... and thanks for your response. I'll follow the links when I get home.

Ben, As with your dof experience, to my (old and somewhat rheumy) eyes there is a non-trivial visible difference in the print rendering at 600dpi compared to 300dpi. I'm not saying that lower resolutions resampled don't look great, but I love the jewel like quality. I first noticed this printing 6 by 4's for family album use from 5D files and was struck by the output, so it's a natural progression really.

Mike

Kinda like 8X10 contact prints! Sounds like fun.

Yep, that's the idea

Actually, I was thinking about this earlier this evening and wonder if the fact that there is real resolution/detail beyond the limit of accuity is something we are already familiar with from the real world - maybe that helps convince us of the richness of the print we are looking at.

Mike

Wow. Just did some testing. Using the same subject as before I tried refocusing vs not and f16 vs f22.

I don't know why but the non refocused was sharper everywhere in the image period. Doesn't make sense as there is a serious focus shift when rotating but I think it's possible the pano program is having to work harder with the refocused images so there is more smearing or something. Have to admit it took some very serious pixel peeping to see!

What didn't take serious pixel peeping is seeing the difference between f22 and f16. I specifically bought the Macro lens because it's reputed to be designed for stop down perfomance but f16 was so very much better than f22 not only for sharpness but overall contrast that it was very apparent even at 50%. Oh well. I think I'll be bracketing apertures when I'm unsure if I have enough DOF but otherwise opening up rather than stopping down.

Ben Rubinstein said:

I don't know why but the non refocused was sharper everywhere in the image period.

Click to expand...

I don't know which blender you used, but some can be 'steered' by using alpha masks. Smartblend can, and it also works with Photoshop CS3.

What didn't take serious pixel peeping is seeing the difference between f22 and f16. I specifically bought the Macro lens because it's reputed to be designed for stop down perfomance but f16 was so very much better than f22 not only for sharpness but overall contrast that it was very apparent even at 50%.

Click to expand...

I'm pretty sure that f/11 would be better still. Diffraction is determined by aperture number, lens design is only a small factor by comparison.

Oh well. I think I'll be bracketing apertures when I'm unsure if I have enough DOF but otherwise opening up rather than stopping down.

Click to expand...

I'd suggest to give focus stacking a try, it really works very well.

Bart

Thing is Bart that when you are both bracketing (exposure) and focus stacking, especially with the amount of focus stacking really needed, it can make each image rather complicated. More to the point, I'm going to have to use focus stacking for some subjects where f22 won't get close to doing the job for front to back focusing, however the moment you get movement such as foliage you're screwed.

The refocused image was just refocused at every rotation of the making of the pano back to the subject. The frames were then made into a pano exactly the same as the regular one and then compared. I wasn't trying any focus stacking.

Next week I'm going out to do this stuff for real, should be fun. After a full 5 weeks shooting weddings abroad I refused to move from the house this week!

Ben Rubinstein said:

Thing is Bart that when you are both bracketing (exposure) and focus stacking, especially with the amount of focus stacking really needed, it can make each image rather complicated. More to the point, I'm going to have to use focus stacking for some subjects where f22 won't get close to doing the job for front to back focusing, however the moment you get movement such as foliage you're screwed.

Click to expand...

As for subject movement, that can be problematic for exposure blending, but focus stacking will only use the sharpest image/pixels.

Bart

OK, just back from trying this in the real world. Some big things learnt.

Firstly I'm composing using a 50mm lens, use the grid lines in my 5D's viewfinder, add a bit above and below and hey presto I have the 6X12 ratio. I then know what the boundries of my image are and stitch accordingly. Except it doesn't work that way. The perspective changes so drastically when making a pano that unless you stay in exactly the same spot as you were when you composed, what you see through the viewfinder is useless! Seriously the perspective changes from that of a 50mm lens to a 20mm and that's one heck of a difference. In the image below to shoot the image with the 50mm lens I used to compose with I would have to walk far behind the tree, just to get everything big enough to fill the frame!

As such the idea about composing and then shooting with my 50mm are out of the window. I compose with the 50mm then not moving attach my 70-200, zoom to about 100mm and start shooting the frames for the pano. That way I get the same perspective. If I move closer then the perspective and 'look' is shot to heck.

That of course brings up some interesting DOF issues, however I found that the foliage is in constant movement, the light moves so fast at sunrise/set you barely have time for the 6-8 frames needed and unless the light is static (it isn't except under full sun and who wants to shoot in that?) and there is no movement in the frame I can forget focus bracketing, to be honest I can forget any bracketing at all, no time. Here in the middle east the sun rises and sets incredibly fast, within 1/2 hour from full sun to dark and you need to work very very quickly. Infact I had to add a 1/2 stop of exposure to my last frame and I only shot 8!

Next thing. I really have to remember to switch IS off on my 70-200 when using a tripod if I want sharpness, was tearing my hair out tonight, couldn't work out why the frames were not bitingly sharp like I'm used to!

I also realised that I need to crop tighter in camera. I was playing with the idea of leaving plenty room for asthetic messing around, I get home and realise that once cropped as I had wanted originally, I'd lost 13 megapixels. Still had 26 which ain't bad but I have to be less sloppy if I don't want to uprez which I don't.

Of course the main problem was the tourists, I could have killed them all, especially the ones who parked their cars on the path (they all got tickets and I really didn't sympathise!) and the school group who sat down on the grass in my composition to the extent that I walked off and shot something else! Seems to be a fact of life shooting in a very touristy area that is also populated.

Last problem was the wind. The Old City of Jerusalem is one big network of wind tunnels through all the winding little alleyways. It isn't the stitching that is the problem, Autopano Pro is incredibly good at that, just 1 minutes after the sun had gone under the horizon I shot this image (waiting for the light to be less directional) and I was shooting a f16 @ iso 400 - 1/3rd of a second. the foliage has a lot, too much movement. I did refocus very slightly as I moved down the image, I had to, it worked perfectly.

Anyway tomorrow I'm going back to do it again, sharper and with no movement this time (I hope!) and next time I'll get the vanishing point right, but so far...

The Tower of David, The Old City Jerusalem.


Apologies again that I'm yet to register the software!

Went back, had the lens on the right setting this time, bluer sky, finally registered the software and hey presto! 30 megapixel file, shooting took under a minute which was good as the light really didn't last any longer than that! I did refocus when I did the lower part, by that time you're shooting stuff far closer to you and the DOF runs out fast. Did the same thing with two shoots today and it worked like a dream.

Hi Ben,

Glad you have been so persistent as it sure has paid off! Also I like the fact that you rendered the final image in B&W and more so spread the tonality well over the composition. The frame is thin and appropriate and the intervening matt is just right to give the photograph a setting a breathing space. Excellent work altogether.

Concerning perspective, that did not change by you altering you field of view. Perspective just depends on the distance to the subject. Now how that might be distorted by a spherical or cylindrical projection in the software, I don't know. Still, by just stitching, AFAIK, the perspective will not change, just if we are at the same proportional distance as the reduction in width of the scene from real life, your eyes would be close up to one part to see as you saw the same angle of view in your position relative to the scene.

Asher

That was my point Asher, to make a pano of several frames of a subject that fits in a single frame using the same lens you need to move closer thereby altering the perspective. With the picture above I framed it using a 50mm lens then shot it with a 100mm without changing position. Because I used the longer lens I needed more frames and hence got more resolution. If I had tried to do the same picture with the 50mm I would have had to change position thereby ruining the perspective and composition that I had seen through the viewfinder originally.

Here are two more:


Approx 30 megapixels each in this final 6X12 ratio, I'm shooting a bit loose to allow for keystoning corrections, etc. 6-9 files each, takes about a minute to do in the field. I'm loving it!

First image was interesting, there was a 8 stop difference between the highlight and the visible underside of the tunnel, I shot it with a 4 stop bracket combined each seperate frame of the stitch in the excellent 'Zero Noise' software and then stitched the results. The resulting image had stupid amounts of DR until I maxed the contrast for effect!

All of these use refocusing and are pin sharp.

And another,

Ben Rubinstein said:

And another,

Click to expand...

So Ben, how is the focus/exposure stacking going, with moving subjects and all?
The 'Zero Noise'software doesn't remove ghosts, if I'm not mistaken.

Bart

It isn't Bart! So far I've not been able to use either except for the 2nd one where there was a 5 stop difference between the underside of the arch and the light beam where I used zero noise. None of them have exposure bracketing apart from that and I've not used any focus stacking, couldn't for those reasons. I've been refocusing as I move along pretty stopped down.

I was wondering on the Zero Noise programs how they worked for you. I had dl'ed them but since they required the ACR I moved away from that because my workflow is much more Nikon centric for conversions. I am still a huge fan of TuFuse and how that program works. Thanks in advance

Mark Prins
www.inanadaimages.com

It doesn't need ACR, it uses DCRAW to develop. It is a problem if you are tied to a specific RAW rendition but I find that the 16 bit TIFF's have so much information I can so whatever I want in PS afterwards.

Couple more to keep you going, all refocused. The first one is sharp from 20cm away from the tripod to the far distance with a 70mm focal length. Impossible to do without refocusing.

Hi, Ben,

Ben Rubinstein said:

As such, should I refocus on the subject each time I rotate making sure that each frame has the optimum focus on the subject or will I start getting problems?

Click to expand...

Note that, depending on the particulars of lens construction, the linear magnification with respect to an object at any specific distance can vary with shifts in the focus state. This can certainly add a complication in trying to stitch the separate frames.

This result can come from two phenomena:

- The basic formula for magnification draws on the focus equation:

m=Q/P

1/P + 1/Q = 1/f

As we focus at a smaller distance (P decreases), for a certain focal length, the distance from the second principal point of the lens to the film/sensor (Q) increases. That increases the magnification.

- In many lenses, as the focus is shifted, the focal length actually changes. Typically, for focus at a smaller distance, the focal length decreases, decreasing the magnification.

Whether the amount of focus shift that would be involved in your situation would lead to a consequential net magnification change I don't know.

Doug, I keep hearing this on forums but in real life it just isn't true, the change in magnification is so very slight that modern software negates it automatically so that everything lines up. It's a non issue as you can see from these examples, sounds good and technical but is inconsequental in the field.

One of the very first things I learnt when stitching is that most of the commonly held beliefs about stitching are very much out of date vis a vis modern software. Quite a bit of close up work here but still yet to need a nodal slide, refocusing works perfectly and you really don't need the setup to be level for the panning. I know older versions of PTGUI couldn't work with refocusing due to the change in magnification and even the present version needs you to decouple the focal length indicator from the images but Autopano Pro does it all automatically and works like a dream.