Tag Archives: Photomerge

Panorama Pursuit

Original Publish Date: 25-May-2011
Last Revision: 06-November-2017

First and foremost I’d like to thank my friend, Eric (Mr. Panorama) Harness for teaching me the basics of this panorama stuff. He’s even helped me salvage a few of my early bad attempts.  Eric knows what he’s doing. I’m a dabbler.

The Scene

15 minutes before sunset clouds and fog over the San Francisco Bay amps up the goldenness of the golden hour. All 24 shots (see below) were taken over a 1 minute and 20 second time span. This view spans NW to North.

What fascinates me about this technique is how well the sun can be included in a normal daylight shot, and how using panoramic techniques is a great way to avoid or eliminate flare. I’m not new to flare. I’ve developed a few techniques to control it. But this method was a pleasant surprise.

The Tricks

There are two important things to get right to pull this off – at least to do so well.

  1. Get the right bracketing of the exposure.
  2. Level that tripod and camera.

Some additional tips that will help a lot include:

  • Use a tripod!
  • Keep the foreground “far away”
  • Use manual exposure settings
  • DO NOT use a polarizing filter
  • Turn off autofocus
  • Rotate around the “nodal point” of the lens
  • Overlap your shots generously
  • Set autoexposure bracketing (I recommend 0,-2,+2 on Canon, or 5 stops on Nikon).

Let me try to explain a little. A sturdy tripod is a must for a stress free panorama and doubly so if you try to use HDR. I don’t have a panorama head, but I discovered that using the tripod collar on my 70-200mm f/4 L lens is a great advantage.

Since things near the camera are more likely to cause stitching errors avoiding elements closer than the hyperfocal distance is a good strategem. Hyperfocal distance is a big word and a slightly murky concept, but I’ve covered it before.

There are holy wars about what the proper name is for the correct pivot point to take images for a panorama (horizontal images) or vertorama (vertical images). I really don’t care what the point is called, but knowing that it is definitely somewhere IN the barrel of the lens is important since the tripod mounting location on the camera itself is definitely NOT in the correct place.  The best camera orientation for taking a panorama is portrait (vertical) mode. Short of investing in a nodal rail and a bunch of other panorama specific hardware it becomes obvious that a ring collar around the barrel of the lens is the simplest “good enough” solution – even if it is not the perfect place.  To find the proper rotation point is pretty straight forward. Have something in the foreground and something far away. Place your camera and the left edge of your frame such that the foreground and background item nearly overlap. Then swing the camera around the attachment point to confirm if at the right edge of the frame the objects have the same spatial relationship.  A more detailed explanation can be found here.

For zoom lenses the proper pivot point may be different at different focal lengths. I merely pass along this observation: using the ring collar at 75 mm on my 70-200mm f/4 L lens I get no stitching errors.  That is good enough for me.

The Setup

I mentioned leveling the tripod. That is important or the panorama will be skewed and detail is lost when I have to crop off a lot of the image – and realism will suffer, too because the horizon will also go weird.  I recommend setting the horizon in the middle of the frame and rotating the camera well past your beginning and ending points and making sure the horizon is level in all the shots. I mess with the tripod and the ball head until it looks good to my eyeball.  I do not use a level or bubble level for this, though I am sure one might help.

While looking at the various frames that will comprise my panorama, I note what the camera metering says. When close to, but not including bright light sources the metering will define the fastest exposure needed. The slowest exposure is defined by the darkest areas.  If 5 stops is not enough, the image may not work. At minimum I want my fastest exposures to be right for the brightest areas (usually the sky) and my slowest exposures to be right for the darkest areas. I choose the exposure that is midway between the extremes as the starting point.

MountAllison_All 1920x1200

In this case, I saw that at ISO 200, f/8 the bright frames metered at 1/1250th while the dark frames metered at 1/80 so I set the exposure to 1/320. Two stops slower is 1/320 × 4 = 1/80. Two stops faster is 1/320 ÷ 4 = 1/1250th.   I took test shots to confirm that at least one of each of the shots in each set was not seriously blowing out pixels.

Shooting starts by setting auto exposure bracketing on, aiming the camera to the left of my intended first target and firing off the bracketed sequence. I then rotate the camera to overlap the last shot by about 1/2 to 1/3 and keep shooting until I have gotten at least one full frame past my intended ending point.

For speed, I also have been producing RAW and small JPG files. Usually I process the small JPGs first. If I later want a larger image, I can start over with the RAW files.

Processing

I use Photomatix Pro in batch mode to process each image set into an HDR image. I found “all defaults” worked pretty well in most cases, but it is possible to tweak things a little if I desire.  Photoshop’s merge to HDR can be used but it seems to be much slower, and I do not know how to automate the batches.  I also find the Photoshop interface to be a bit clunkier.

Once each of the HDR images has been created the HDR images can be loaded into Photoshop and stitched using the “photomerge” operation. When I used the Photoshop photomerge operation it left some serious flare in the image which I had to manually fix up by adjusting the layer masks. By contrast Microsoft ICE – a free tool – removed the flare automatically by using the non-flared overlapping portion of another image.

I finished the image in Picasa by adjusting the white balance (cooler), adding tags, downsizing and adding watermarks.

The same evening I made the featured image, I also created this image.
Wide Open Spaces [C_031244-61]

I much like it, too.

Thanks for reading. Comments are welcome. Sharing with your friends is encouraged.  And if you are further interested, please join me at one of our Workshops where you can learn this topic and many others.

Oh, and keep an eye out for Improving your Panoramas written by Eric Harness, yes, THE Eric Harness who taught me the technique in the first place.

Stitching Stars

Every once in a while, I have to try something a little “left field”.

What if, I thought, instead of stitching a panorama or vertorama of a landscape, I tried stitching a 180 degree vertorama of stars.  I have stitched a landscape shot with a star shot before as shown in the image “Above and Below”. For “Above and Below” I combined a stack of star trails with a shot from just before dark of the landscape. Unfortunately after stacking the star trail for the top, I found none of my image tools could be convinced to stitch the portions of the image together satisfactorily and the two halves of the whole lay dormant in the bit bucket bin.

Above and Below

The star shot – that is the top half – is very similar to this shot on Flickr. I opted to use a stack with fewer frames to keep the sky dark and let the smear of the Milky Way stand out. I then very crudely combined the two shots using the “Collage” tool in Picasa3. I followed the crude paste up job with the “touch up” tool to blend the seam lines. The touch up tool is a crude approximation of the Photoshop Healing Tool and Clone tool combined into one. Sometimes it is very effective, sometimes not so much.

But one evening I found myself staring at the dark skies at 9,000 feet above the town of Bishop, California. The Milky Way stretched from horizon to horizon and was nearly directly overhead. So I thought I’d try something that I expected to be difficult… but proved surprisingly easy. I thought I would try to stitch together the stars in the night sky in much the same way that one can stitch a panorama together.

But Would Stitching Stars Work?

To answer the question, I took the following 8 shots at high ISO, 30 seconds each.

I then dropped them into Photoshop CS5’s PhotoMerge tool and got this peculiar and clearly poor result:

Photoshop’s confusion was partially understandable.  When I took the eight shots for the image I shot 4 shots from the horizon up to the zenith. I then turned the camera 180 degrees and shot the remaining 4 shots descending from the zenith to the horizon.  The top of the first shot must be stitched with the bottom of the second shot, the top of the second with the bottom of the third, the top of the third with the bottom of the fourth … and then the great discontinuity: the top of fourth shot needed to be stitched with the TOP of the fifth shot.  Realizing that the problem might be the topsy turvy issue, I rotated shots 5 through 8 to preserve the “top to bottom” alignment. Unfortunately using the default mode Photoshop could not figure out what to do with the topsy turvy or the consistently aligned images.  So I tried another approach. I changed the Photomerge projection mode to “cylindrical” and obtained the following:

There is a peculiar and inexplicable bulge in the result, but it certainly looks quite a  bit more realistic than the first try.  It took about 5 minutes to do the stitching.

To compare, I fired up the Microsoft ICE tool (an Image Composite Editor – free from Microsoft for PC users) and dragged the 8 original un-rotated exposures and got this:

in about 3 minutes. The ICE result is a lot closer to what I was expecting and the tool did not seem to object to having either the topsy-turvy or properly aligned images thrown at it.

After cropping the ICE result and a little image clean up I arrived at the following visual conundrum. Which way is up? Well, the middle of the picture is up! The top and bottom are the East and West horizons.

Path of the Milky Way West-to-East [C_009575-82stitch]