🚀 go-pugleaf

I stitched all five separate focus layers as full panoramas and then combined them manually, Larry. That's why I chose this location as it gave four distinctly separate zones with minimal overlap; the masking was at pixel level and was extremely complex, as you'd imagine. The focus at the "macro" parts at right was the biggest hurdle as there was no real defining edge between the first and second layer of focus, the division of which ran right through the middle of the moss.

All the photos had to be done in situ at the one time in order that the out of focus areas behind bled the right colours into the edges of the sharp layer above, and allowing the sharp version of the under layer to be dropped in unobtrusively. Repeat x 5, which is why it took so long.

The senator himself took three frames of the pano of his layer, so not only did he have to sit still for each pass, but it was absolutely imperative to have the detents available in the 302 head to enable rapid transition to consecutive shots accurately.

I desaturated the background layer a bit to add to the "spooky" feel and separate the main subject a bit more, and to help disguise the colour/contrast shift that happened when the sun came out for the background sweep.

When I say 35 photographs, of course that was the end use - I took around 500 shots altogether, not counting the four trial sessions over the days prior to the final shoot. It was not something that could be repeated, so I had to go with what I could get on the day and that meant being as prepared as possible beforehand. The good senator alloted me two hours of his time, thankfully being a keen and accomplished photographer himself and so able to understand the difficulties of what I was doing, and the need for him to actually come out to the forest in person and not just be stripped in later.

I stitched all five separate focus layers as full panoramas and then combined them manually, Larry. That's why I chose this location as it gave four distinctly separate zones with minimal overlap; the masking was at pixel level and was extremely complex, as you'd imagine. The focus at the "macro" parts at right was the biggest hurdle as there was no real defining edge between the first and second layer of focus, the division of which ran right through the middle of the moss.

When I say 35 photographs, of course that was the end use - I took around 500 shots altogether, not counting the four trial sessions over the days prior to the final shoot. It was not something that could be repeated, so I had to go with what I could get on the day and that meant being as prepared as possible beforehand. The good senator alloted me two hours of his time, thankfully being a keen and accomplished photographer himself and so able to understand the difficulties of what I was doing, and the need for him to actually come out to the forest in person and not just be stripped in later.

You certainly chose the most complicated path in doing the layers first as full panos, Fred. If you got just one layer out od sync....

Is there a difference in combining all the layers first, then doing the pano from flattened layers?

How do you deal with the slight motion of plant life at the near macro or macro level in such attempts? Particularly in the focus layers?

I've found that combining first and then stitching leads to distortions between frames that the stitcher can't resolve with its algorithms, which I guess are based on a supposed input and could not take into account a stacked focus blend.

Doing the stitch first and then combining is more accurate as the 302 pano head stops in the same position at each frame via the detents, so each focus layer of the stitched pano is similar in registration. A bit of image scaling to match is sometimes necessary with the individual layers, but the match is usually close enough to avoid major dramas, just a resize for the shift in focus mainly being necessary.

I have also found, as you pointed out, that to get few artifacts with CZM or similar auto stackers you have to photograph many small shifts of focus over a limited range - go too far (as this photo would have needed) and you get out-of-focus auras appearing around sharp areas. Sometimes manual really is the only way.

Movement isn't such a problem in the macro side as it takes a fair breeze to get these small things moving - the biggest problem is the large fern fronds which move at the slightest whisper of air movement (as happened with the frond in middle-distance to the left of the photo). Given a choice I would not have photographed on that day - as soon as the sun came out the wind picked up. I wanted to do the shot in June, which is usually cold and still, but the Senator's parliamentary sessions didn't allow the time then.

I was thinking of the ferns.

I did a fern shot last December...with lots of snow on them. They didn't move. I could actually take a pair to stitch.

Using the grid on the screen of the Nikon as the intersection point results in a near perfect to perfect square for a double frame verticals, for which I am delighted. Figuring out the composition as such is a bit tricky, but sometimes the surprise resulting is better than if I had done it with the Hasselblad.

When i look at that pano, it does not strike me as a shot that needed more than three frames to cover that view. why the necessity for such a huge number of shots if i might inquire?

Look at the finished size as a first approximation, Donald. I've used five frames for less coverage than Fred did.

It covers about 180 degrees horizontal, about 70 degrees vertical, using 7 vertically orientated shots. The camera was about 18" from the Senator's face.

Choose your subject carefully and the actual angle of view becomes very deceptive. The tree at left was in a line slightly behind my left shoulder, the moss at right just in front of my right hand when I was facing the tree fern behind the Senator's right shoulder. Overlap was about 20% at each frame edge.

And speaking of panos, don't forget the pano thread:

 <http://www.adobeforums.com/webx/.1de5f905.59b6ea93/52?14@@>

Thanks fred, that is surprising to hear that wide. What is the main reason for overlapping so much, and not getting just the edges to overlap?

Most stitching programs like to have at least 15% overlap. This becomes particularly important when you use wide-angle lenses as the stitcher warps the image at the edges to take account of the wide-angle edge distortion. The more overlap you give, the better the chance for a 100% registration at some part of the overlap.

I always seek to avoid as much as possible retouching double-image parts of a panorama as this can be incredibly time-consuming for no benefit other than to rectify something that was avoidable in the first place.

I was using only a moderately wide 35mm for this shot, mainly to keep the wide-angle perspective distortion on the Senator's face to a minimum, but had I been using a wider 24 or 22mm setting I would have allowed 30% overlap at least.

Again, this is another reason for using a good pano head as the Manfrotto 302 (303) has adjustable detents which keeps the overlap consistent from frame to frame and gives the stitcher a better chance of matching the whole thing a lot better.

Fred, how much time was there between the exposures that included the Senator's face? However much it was, he did a good job of holding still!

Also, Fred, I'd like to add that this is the kind of posting that I look forward to here. Good information, presented by an excellent photographer and teacher.

Thanks, Nick.

I lined up the whole shot so that Bob's face filled the center of a frame, then shot the series. My main instruction to him was to not move his shoulders, as they were on the frames left and right of the face frame. The detents on the 302 head mean that you can swing from one to the next accurately in under half a second, which makes people panos practical, though with the slow exposure in this case I had to let the vibrations settle before tripping the shutter.

As mentioned, Bob is a keen and accomplished photographer himself (he's often said that had his environmental politics and activism not so consumed his life, he would have been a full-time photographer), so he grasped pretty well immediately what I was doing and why, which made the shooting a lot easier than it could have been! I'm going to an opening of his latest exhibition tonight as it happens, though how he gets time to take photographs these days is beyond me.

Thanks Fred. can i ask one more favor of you? in post 18, I displayed a set of three pictures demonstrating a test to find the nodal point, but even though the parts have corresponding sections that seem identical, if i slice the three shots and try to join them, it still does not give a seamless pano like i think it should. how would you explain that visual discrepancy?

That's because of the perspective distortion of a wide lens, or because of the different POV when swinging the camera with something close to the lens as well as something more distant in shot. The important thing is the relationship of the near and distant parts to each other at left and right of the different frames so that their vertical alignment is the same.

The stitching program will warp the hroizontal lines to fit left to right, which will end up in a horizontal line that extends across the whole image taking on an approaching/receding bend as would be seen with a rotating pano camera.

Also, a stitching progam like pano tools will endeavor to obtain the information about your lens, either manually or EXIF, if available, and compensate fo that lens, magnification etc.

One other question Fred, and that is focus. Autofocus would be a joke so it has to be manual. Do you set the near/far points by eye then incrementally move the focus ring through that range? For WA lenses, it ain't very far!

I did a test recently to check the wisdom about manually setting the focus point and shoot all frames at that setting so as to not change the image size. So I set up a test with a row of trees on the diagonal with the farthest being 100' and the nearest 10', and each time I moved the camera I refocused. Stitching the set proved to be a cakewalk and the final presentation has great DOF from the left to the right.

I've shifted focus during the sequence in previous panos when required (Pixentral is down for me at the moment but I posted a pano of a wood-stave pipeline that was done like that a while back), but in the case of this layered focus pano I focused manually on the immediate foreground first, then shifted focus back a bit, then focused on the Senator, then the tree fern behind him, then the background, doing a complete 7-shot pass each time, and doing several passes of each layer for security as well.

Of course each layer then had much of its area out of focus, but by combining just the focused bits of each I got close to infinite depth of focus overall. In the final version the background at left is from a different pass to that at right.

Someone has nominated the term "constructed reality" for this sort of thing - it looks convincingly like a single image but although everything is shot during a short time period and at the same place, the actual scene never existed together as portrayed in its entirety, and in a way far beyond that of a normal stitched panorama. For instance the Senator was off photographing mosses and ferns for himself while I was doing the background passes.

I just purchased a new tripod. Why? Because it's perfect for panoramas. The Manfrotto 755XB (those with extra bux might want to look into the carbon fiber version) It is a video tripod, and it's usefulness lies in the half sphere head on the center column. Uneven ground? No need to worry amount a level platform. The platform can now be leveled using the column adjustment. Easy peasy! Removed the big gumption factor in taking the time to set up the tripod... leveling with the damn legs.

I did a test, and simply leveling the platform with the built in level provided an almost perfect edge alignment. No more stair steps!

I wish it had even more precision in leveling but it's so close as to be far more useful than leveling with the legs.

Thought this would fit in this thread moreso than any other. 300 one minute exposures.

 <http://antwrp.gsfc.nasa.gov/apod/ap090314.html>

Fascinating � including the descriptive text. Thank you so much for posting the Link.

I'm curious as to why he tripped the shutter over 300 times for a minute at a time instead of just leaving the shutter open? I can't see how the end result would have been any different other than avoiding a stitching-from-hell exercise ;-) .

Seeing this, I reckon pointing a full-circle fisheye lens straight up at the south or north pole and exposing for 24 hours in mid-winter would get the ultimate star trail picture. All lines should be continuous circling around the center of frame with a band of "sunset" 360 degrees around the horizon.

Maybe he doesn't have a cable realease...

      exposing for 24 hours in mid-winter




I can see at least two disadvantages to that idea: winter temperatures are about -70�F (or less); and Long-exposure digital Noise would be quite a factor also!

:(

That noise level would be less at -70, however, the camera would need some warming as -70 would also have a detrimental effect on the batteries.

I;m also not sure about the silicon at that temp. Nor anything mechanical

If they could put a man on the moon in 1969 (temperature fluctuations there can make the arctic look like Palm Springs and the Sahara look like winter in Siberia), I'm sure someone in 2009 could work out a simple way to keep a camera warm....

If digital noise is a problem, why not use film? (how soon we forget....)

Some times even well-known authors using telescopes and digital capture devices confuse how signal to noise increases with increased exposure and come to believe that 300 one-minute exposures equate to one 30 minute exposure (ten to one ratio).

Ouch! That's wrong.

Signal to noise increases in proportion to the square root of the exposure whether this is a single exposure or a stack of multiple exposures.

The single exposure is actually better, because it has a single read-out noise component, whereas the 100 one-minute ones will have 100 read-out noise components.

Of course, DSLRs are not cooled devices, so you don't want to leave the shutter open for hours. The sensor gets really, really hot.

I've been doing multiple night-time exposures of the night sky for a long, long time. When shooting deep sky (galaxies, nebulas and so on) digitally, I try keep each exposure to less than 240 seconds, occasionally even 30 seconds, usually on a motorized equatorial mount.

One thing to remember is to switch off auto noise reduction (dark-frame subtraction), if for no other reason than to save imaging time.

The image below was taken with a 300mm Pentax manual lens on a Pentax *istD, in Western Oregon a few years ago. Dozens of shots were stacked, I'd have to dig my notes to be sure, but I seem to remember it was in the order of 70 or 80 stacked exposures. I used a Canon cable release with an intervalometer, adapting it for Pentax contacts.

The sensor/chip in the *istD, which was also used by Nikon in their first DSLRs, just happen to be the one also used by the the Starlight Xpress dedicated cooled CCD camera, which sold for about $8,000 US Cy. at the time. In those days, Nikon had trouble with noise reduction that could not be switched off which deleted faint stars. Bad karma. ;) Canon was not in the running for those purposes yet.

 <http://www.pixentral.com/show.php?picture=1Bsn1LuIuXj1XL8Tkc48qpwaskIBb0>

On keeping the camera from freezing during a long exposure.....

      The sensor gets really, really hot.




Well, that was simple ;-)

Ouch, -70 is kind of chilly.

I took an hour exposure in 10� weather the January before last. But then I was lost and waiting for the moon to come up so I could find my way back to the trail so had nothing better to do.

The single exposure is actually better, because it has a single read-out noise component, whereas the 300 one-minute ones will have 300 read-out noise components.

Not quite right. It depends on how you use the multiple exposures. If you combine them in a summing device, the s/n ratio goes up as a function of the number of images being summed. It is a log function, not a square root function in that case.

Of course, the closer you come to the exposure threshold, the lower the s/n until the noise swamps the signal.

CCDs have inherent noise characteristics that makes them the better candidate for sensors than CMOS. But CMOS is far cheaper to manufacture, as CMOS is so ubiquitous in silicon production, and it's cheaper to deal with the noise using software.

A Peltier cooler is indicated if the chip is indeed heating up to the degree you indicate, Ramon. In any case, thermal noise is only one of the sources for sensor noise.

Fred, the only films that can even come close to today's sensors is Tech Pan and Kodachrome, With K a distant second to Tech Pan.

IMO, of course. :-)
 Nice shot of Andromeda Galaxy, Ramon.

      Fred, the only films that can even come close to today's sensors is Tech
      Pan and Kodachrome, With K a distant second to Tech Pan.




I know, Larry, but if the temperature is too low to allow a digital camera and its batteries to function, a Nikon FM2 (or any other fully mechanical camera) with an 8mm full circle fisheye attached would romp the project in (even if one had to open the shutter in a warm place first and control the actual exposure with a lenscap should the lubricants freeze and seize the shutter).

I wasn't suggesting that this should be what we regress to, merely responding to Ann's observation of noise occurring in very lengthy exposures in digital cameras. I haven't yet tried an exposure running to hours to confirm that this indeed will happen with a D3, only commenting on that supposition as put.

Gotcha, Fred. :-)

Actually, I had that experience at only 15F. The Hasselblad worked, the 20D didn't. (Not mine, a friends)

That is a great shot Ramon. Thats one area of photography i wish i had the equipment and knowledge to try.

Thanks, Donald. You already have most of the equipment, the only thing you might lack is a motorized equatorial mount. Contact a local astronomy or astrophotography club. I'm sure a nice fellow like you can borrow one. :)

Another thought, Donald: if you can run any Windows emulation software on your machine, you can use the free Iris software program, which incredibly gives you a virtual equatorial mount. It's mind boggling.

It's also stable enough that I've managed to run it in SoftWindows98 emulation software.

EDITED last post to insert live Iris <http://www.astrosurf.com/buil/us/iris/iris.htm> link.

Not ingnoring you, Lawrence. :) Just researching your comment re 

      in a summing device, the s/n ratio goes up as a function of the number
      of images being summed. It is a log function, not a square root function
      in that case.

It's a technique that goes back to the 50's in the detection of nuclear events in a lab, the co-incidence detector. I have used it to make a super noiseless image. The Nikon will do it for you in the camera software, and you have to know what you are doing when you try it. Nikon calls it Multiple exposure and is meant for such things as catching several bursts of fireworks, for instance. As soon as I saw it, I knew it would operate in the coincidence detector mode. The noise is random, the image is not.

You can adjust the gain in the multiple exposure as well for more control of what yo seek in the end

Just to clarify, I'm not talking about multiple exposures in the camera at all. I talking about individual images produced discretely, one by one, and stacked in the computer.

The D3 camera takes individual photos then stacks them, digital cameras cannot really successfully take multiple exposure as is traditionally known when you capture it on film on one frame. The noise level for doing one extra exposure would be off the chart I understand.

The images produced in camera are also discreetly produced. Then they are stacked so that the total output for the information part = 1. So, if you shoot, say 4 shots, each shot generates 1/4 the information.

The key is that the information is coincident, that is, there is no variation between the information content except for the amplitude. If there is any difference at all, that difference will be seen as noise.

When C3 Extended came out, there was a routine by which one could do a series of shots with fixed camera location but certain variations such as cars or people or ??? which do not repeat from frame to frame. Summing all those frames in the tool provided would magically remove those elements, but not perfectly. That is the process. When you are stacking the star photos, I don't think you want to remove anything at all from the frames because you are exposing over a pretty long time frame and the individual frames are not identical. When I want to remove noise, the noise is the element that differs frame to frame, in a Gaussian distribution (hopefully!), but the information, say a building,is identical. Obviously, you need a firm platform so I only try this with tripod shots. I'm sure that an algorithm can be derived to do this with hand held images but the noise reduction would suffer.

Essentially, this is an analog process, so the digital implementation has to account for that.

The key to understanding is to ask yourself: What is signal (information) and what is noise, and how do I maximize information? Indeed, how do I know at all what is and if there is information in the data stream.

In the original Star Trek series, there was an episode where the crew had to deal with a hidden renegade, and could not be found with conventional means. So they contrived to set the sensors to read heart beats, and the crew had each of their heart patterns recorded, so that they could be nulled out ie, theirs were "noise". Eventually, only one beat was left and voila! (or is it viola? Ramon?? )

      images produced in camera are also discreetly produced




:D

Voil�!

discreetly

dis�creet (d-skrt)
adj.
1. Marked by, exercising, or showing prudence and wise self-restraint in speech and behavior; circumspect.
2. Free from ostentation or pretension; modest.
[Middle English, from Old French discret, from Medieval Latin discrtus, from Latin, past participle of discernere, to separate, discern; see discern.]
dis�creetly adv.
dis�creetness n.

Wade, 

      The D3 camera takes individual photos then stacks them




Up to how many? ;)

Hmmm, the spell check flagged it that way! I didn't check the spell check.
Discrete is what I intended!

But then there is also Descrete.

descrete
Discrete
Descreete
Discreet

All valid terms.

I think 7 or 9. I forget.

I haven't check the D90 yet either.

When I get home....

      All valid terms




Each with a different meaning. :)

Wade, 

      think 7 or 9. I forget.




Thanks for that info. Of course, that's not really in the vicinity of what's required for deep sky shots with a 300mm regular telephoto�as opposed to huge astronomical telescopes.