Hi Folks:
Making digital panoramas is essentially a two-part process. In part one of this post I covered a bit about digital panoramas in general and some considerations that become important when collecting the images to be used for the panorama. Part two is focused (pun intended) more toward what to do with the images once you have them on your computer.
NB: If you have a smart phone you can use the panorama mode on your phone to make a simple panorama. Some even allow you to create a panoramic image from a video. Depending on your phone and your expectations, that may be sufficient for your needs. For me, it’s mostly not, because one of the benefits I find in making a digital panorama is the increase in resolution I obtain from joining together several images into one. The downside to that is that file sizes can get quite large, so when rendering the final image file it’s best to balance what you want against the capabilities of your computer. Sometimes I try to make smoke come out of mine… 🙂
This post is (typically) very long, and so we’ve broken it up into segments for you. Clicking on the subtitles will bring you to the relevant section:
How Panorama Software Works
Projections
Panorama Software Options
Making Panoramas
Stitching Errors
Exposure and Image Noise
Parallax
White Balance
Chromatic Aberration, Fringing Colour Artifacts
Image Cropping
Keystoning
Final Thoughts
Before I continue, in part one I mentioned that (technical cameras and tilt/shift lenses aside) unless one is using a panorama head mounted to a tripod (the best option) it’s important to rotate the camera around the entrance pupil (some people call it the nodal point, but that’s different) of the lens to avoid parallax errors. If shooting handheld, the best way to achieve this is to move the photographer rather than the camera. If you’re using a smart phone to make the base images (my Samsung Galaxy S21 shoots .dng files in Pro Mode), this task is relatively simple. Place the index finger of the hand not holding the camera on the edge of the phone, in line with the lens, and rotate the phone around that point. Remember to keep your fingers out of the frame. If making a really wide panorama, keep the camera as stationary as possible, and move your feet. If using a DSLR or CSC camera, you can either place your finger under the lens (if you want to get technical you can look up the specifics of just about any lens online) about where the entrance pupil is and rotate the camera around that point. My Laowa 12mm lens has the entrance pupil clearly marked on the lens barrel. Depending on your foreground distance (where parallax is more of an issue) or if you’re using a P&S camera for example, take something small like a coin, a pine cone, a stone… and place it on the ground in front of you. Imagine your camera suspended above this point, and rotate your camera around that object. Again, move your feet rather than the camera. Stitching software will do its best to accommodate you.
How Panorama Software Works
Okay, to understand how panorama software works, find something rectangular and fairly flat – it could be your phone/ tablet, a book or even a piece of heavyweight paper. Holding the object in both hands, try rotating one side of the object toward you and away, tilting the top forward and back, then tilting the top to one side and the other. Each of those is known as an axis of rotation. Aircraft pilots (and panorama photographers) refer to them as yaw, pitch and roll. In making the base images for a digital panorama you’ll mostly be either rotating the camera from side to side (yaw) or tilting the top of the camera up/down (pitch). Roll – tilting the top of the camera to one side of the other is best avoided as it leads to tilted horizons. This can be corrected in software, but I find it best to get as much right in camera as possible. More on that later.
(Click on any image to see it larger)
Now, to connect one image to the next in building the panorama, the software will establish Control Points. Basically, a control point is a point that is visible in two or more images. It could be a rock, the edge of a tree branch, a corner of a building… anything identifiable as the same place in two (or more) images.This image shows the Control Point Editor in PTGui. In this case there are 8 images making up this panorama and this image shows #1 and #2. Each control point is assigned a number and the same number appears in both frames. At the top, highlighted in green, one can see that image #1 has matching control points in images 2, 3, 4, 5 but not images 6, 7, 8. This highlights another important consideration in making base images: overlap. In order to link two images together, there must be an overlap between them so that both images can display the same control points. The general rule of thumb is 30% overlap; more overlap is better, but also leads to having more images to link up to make the panorama. Less overlap – say 10% leads to less confidence in having the images line up. For a specific case of this, you can read: Panoramas: Making Waves. I’ll wait…
Welcome back. Now, each set of control points has a number as well as a distance. It’s easiest to think of the distance as the confidence the software has that the control points in each image are properly attributed…the higher the distance, the less confidence. This could be because of a moving object (wind or waves), because of parallax, or because of excessive camera movement. Remember your movements of the phone/book or whatever, being able to move in three axes of rotation. In a perfect world each image would be lined up like dominoes, varying on only one axis, but if shooting handheld that’s not often the case and the software does its best to adapt. NB: To the best of my knowledge, all panorama software uses control points. However, not all software shows you and/or gives you the option to edit/ add/ delete control points.
Projections
This leads us to the next stage of stitching a panorama: projections. Once the control points are set, the software will do its best to align all of the images together based on matching control points. In the film days we called this a photomosaic, and photomosaics were made by using a pile of prints, a pair of scissors and some glue. Parts of each image were cut out and overlaid to make the final pattern, aligning edges as best possible.
This image is from Holland Point, Victoria, BC, looking west toward Ogden Point and the Juan de Fuca Strait. This is a 6-image stitch; the red lines show which parts of each image have been combined to make the final image.
At the risk of overstating the obvious, every image we make requires us to take a 3-dimensional subject and capture it in a 2-dimensional space. We compensate for that as best we can using different focal length lenses, using Depth of Field creatively, etc. Once we get beyond a single image, we raise that to another level. Imagine this: Take an orange or a grapefruit or equivalent, place it on a flat surface, put something flat on top of it and press down. Hard. Not only will the result be messy, but it’s clear that there’s no way to flatten that sphere without making some tears along the way. Sacrifices have to be made, and the ways the software bends and twists the data will lead to different distortions.
Each way of bending the data to make a panorama is called a projection. Different software will offer different projection options and some more than others. The three most common are planar (rectilinear), cylindrical and spherical.
Planar, as the name implies, suggests that the aligned images are flattened as if on a table. Depending on the scope of the panorama, this can work well, but on longer panoramas there will be more distortion out toward the edges.
A cylindrical projection implies that each image is placed around a cylinder (think of something like a toilet paper roll, which is then slit and flattened.) Cylindrical projections tend to have distortions that round straight lines. In the image above there is less distortion at the back and edges of the image but the front edge of the flower garden should be straight.
To me, the simplest way to imagine a spherical projection is to imagine a soccer ball, with one image occupying each of the pieces sewn together to make the ball. Spherical projections are most often used with multi-row panoramas, but the distortions are largely the same as with the cylindrical projection.
Panorama Software Options
Now, how many projection options you have to work with is varied and entirely dependent on the panorama software you’re using. If I remember correctly (it’s been a long time), Lightroom has 4, Capture One Pro has 4, PTGui has 13, and so on. This means it’s probably a good time to talk about stitching software. Over the years I’ve worked with several panorama stitching programs and I’ll start by mentioning that in my experience every panorama stitching program will provide slightly different results – even with the same images and using the same projection. The algorithms for each are different. In alphabetical order:
Adobe Photoshop Lightroom – I used Lightroom from Beta 1 to ver. 6.14, which was the last standalone version. I moved to Capture One Pro when Adobe started renting software exclusively. ’nuff said about that.
Affinity Photo – currently at ver. 2.5.5. Affinity Photo’s panorama option is built into the software and is a good, basic option. It does not provide different projection options when stitching.
Autopano Pro – no longer available, but I still have an older version. It was, to me, the most powerful stitching program I’d used; unfortunately the older version I have doesn’t work well with modern raw files in terms of colour. I still use it for .jpg files, and sometimes I’ll use it with .tif files rendered from raw files.
Capture One Pro – currently at ver. 16.4.5. Capture One is my go-to raw conversion and image processing software. The panorama stitching is basic (four projection options) but for much of my work more than adequate. NB: Capture One Pro will only stitch raw files, not rendered files (.jpg or .tif for example).
Hugin – open source, free software, updated yearly. Hugin is very powerful, but in my experience the UI (user interface) might be a little frightening for beginners. Give it a try and see for yourself!
Microsoft Image Composite Editor – no longer available. I played with it a bit but never used it much.
PanoramaStudio Pro – currently at ver. 4 It’s good software, and I used it for a few years before moving to PTGui.
PTGui Pro – currently at ver. 12.26. PTGui is probably the most powerful stitching software available today. It’s also by far the most expensive. There are two versions, a standard and a pro version, each offering different options at different price points.
There are likely other stitching programs available on the market, but these eight are the ones with which I have experience. Which one you choose depends on your needs, budget and other factors such as what raw converter you’re currently using and whether or not stitching is built into that.
Making Panoramas
Okay… you’ve made your images, uploaded your files to your computer, and now you have some decisions to make. The first step (in my opinion) is to group together the images for each panorama. I’ve made panoramas from two images, and I’ve made panoramas with over a hundred images. When you’ve just uploaded a thousand images or more to your computer, it can sometimes be challenging to separate one set of images from another. What I (almost?) always do when making panorama sets is to photograph my finger at the end of a set in order to separate each set. One can also check the time stamps for each image as making panoramas involves making a series of exposures in a relatively tight sequence, followed by setting up for the next and/or making single images. Orientation changes (portrait vs. landscape) can also mark the end of one set.
Once you’ve segregated your images (I find colour tags work well), the next step is to look at each image in the sequence and check to see if it’s in focus. A blurry image in the middle of a panorama set will not work. When making a sequence of images, also make sure that your shutter speed is high enough and/or you allow enough time between images to prevent motion blur when making your exposures and when moving between them.. This is when those things come back to haunt you. If you’re shooting handheld and your panorama includes a visible horizon, this is also where you check to make sure the horizon is level in each image. Stitching software can go to some extent to compensate for this, but you can end up with something that looks like this:
As you can see, the horizon in the first image is nearly level, but the following images have increasing amounts of tilt (roll). There are a couple of ways to deal with this, depending on the software you have. One option is to use your raw processor (again, I use Capture One Pro) to straighten the horizon in each image before stitching them. However, if you’re going to then pass those images onto another program to stitch them, you’re going to have to render them out as .jpg or .tif files in order to stitch them with the roll corrected. If you’re using your raw processor to stitch the images, it will use the corrected raw images to stitch.
If I take those seven images as is and run them through PTGui, I end up with this:
Not too good. If you want to continue working with this image, you have a couple of options – again depending on the software you have. If you have a pixel editor like Affinity Photo you can use the Mesh Warp tool to drag the edges of the images down into place. NB: What I do in a case like this is to first create a new layer, and on that layer use the Pen tool to create a horizontal red line in line with the horizon (hold down the Shift key to keep the line level). This gives me a point of orientation when manipulating the underlying layer.
With PTGui, the software has tools to manually adjust the yaw, pitch and roll of the aligned images. By changing those, I end up with this:
Much better.
Stitching Errors
Another thing you can encounter are what are known as stitching errors. Stitching errors are often caused when one or more sets of Control Points are off. This is often caused (again, when without a panorama head) if one accidentally tilts the camera forward or backward between frames or if the foreground is close and you have parallax errors. Here’s an example:
This is a two-image stitch, and as can clearly be seen, there’s a difference in the pitch of the camera between the two frames. This has caused a break in the horizon that’s pretty easy to see. When dealing with stitching errors like this, one can try to add/edit control points to bring things into balance, one can try to use layer masks and selections in a pixel editor like Affinity Photo to move a part of the image, but mostly when I get stitching errors I mentally kick myself, delete the files and move on. In this case there isn’t enough detail in the background to adequately create new control points.
On with the show. As I mentioned in part one of this post, I’m assuming that you’re working with raw files, but most of this information also works with .jpg files. The difference between the two in terms of the amount of information available with each image is enormous, but it comes down to what equipment you have and what your final goals are for each image. We’ve covered projections, software options, image corrections (to a degree) and stitching errors. To reiterate, when making the base images for your panorama you want to be as consistent as possible. This means locking the white balance wherever possible. Do not use polarizing filters. It means using manual exposure mode to have lighting consistency. It means – as much as possible – only moving the camera along one axis of rotation at a time when making the base images. It means using hyperfocal distance or a fixed focus distance for each image. All of these contribute to the success of your final stitched image.
Exposure and Image Noise
There are a few things we have yet to cover. One is image noise; another is exposure. These are both very important when making panoramas in the rainforest because the amount of available light from the sky to the forest floor is highly variable. I mentioned this in part one of this post, but the ISO you choose when making your images will impact several things. The first is dynamic range. Again, dynamic range is not the brightest or the darkest areas in your image but the ability of your camera to separate them. You may claim that your camera has ___ stops of dynamic range, but remember that this is at base ISO. For example, my Sony a7RIII has a photographic dynamic range of 11.63 at ISO 100, but half that (6.37) at ISO 6400. Lower ISOs mean longer shutter speeds, and in the darkness of the rainforest that generally means using a tripod, mirror lockup (on DSLRs) and remote shutter release to eliminate camera shake during exposure. Higher ISO also means more noise, and again, this is camera dependent. When reviewing your images before stitching, there are choices to make regarding exposure adjustment, noise reduction, etc. before or after stitching the images together. Rather than reiterating that here, I’ll refer you again to our post Low Light, High Noise and ISO Invariance. Again, choices made when capturing the image data impact the steps to take when making the panorama image.
I’ve mentioned noise issues and lighting variations and the challenges they can bring to post processing. Depending on your camera and lens there can be others. With my camera I find that the range of lighting can yield a stitched image that has way too much contrast, and the only way I can go forward is to take the Contrast slider way down before I can proceed. I’ll usually bring the contrast back up some, either using the Contrast slider, Shadows/Highlights or even Dehaze, but it gives me a place to start.
Parallax
Now, all of the images above were examples made with my Samsung Galaxy S21 phone. From this point we’re going to be working with images from my Sony a7RIII. We covered what parallax is in part one, but if you missed it, I refer you again to this Wikipedia article on parallax error.
Again (at the risk of overstating the obvious), in open country stitched panoramas can be truly panoramic. This is a 16-image stitch:
Black Mountain country, Kelowna, BC
In the rainforest, however, things are decidedly different as the sheer amount of growth brings everything into closer focus. This is a four-image stitch.
World’s Gnarliest Tree – Upper Avatar Grove, Port Renfrew, BC
Because of this, parallax is much more of an issue when making panoramas in the rainforest and more care must be taken when capturing base images. For much of my work, a single tree is often the focus, whether making a vertical or a horizontal stitch. Further, because of the close proximity of the subject, I find that I often have to make a multi-row panorama in order to fit the entire subject into the frame. This image, for example is a 24-image stitch (four rows of six images):
Resting Place – Pacific Rim National Park Reserve, Tofino, BC
This is where poor camera technique will come back to haunt you in terms of parallax and stitching errors.
White Balance
Because of the range of lighting from the tops of the trees to the forest floor, white balance can also be a challenge. It’s best to set a manual white balance before making the exposures, and to do that it’s handy to have a spectrally-neutral gray card or something like the Calibrite Color Checker Passport. If you don’t have either a gray card or a Passport, here’s a tip from (if I remember correctly) Peter Eastway (if it was someone else, my apologies!)
In your raw processor/ image editor, first set the White Balance to neutral (5000K, 0 Tint). Now, crank the Saturation slider way up (say 75 or 80%). Now, slowly move the colour temperature slider to the left until it looks weird, then move it back to the right until you’ve gone too far. Continue moving it slowly back and forth until it looks about right. Now, move the Tint slider slowly to the left until it looks weird, then move it to the right, etc. Keep moving the Temperature and Tint sliders until the image looks good to you. Set the Saturation slider back to 0. The first thing you’ll ask yourself is, “What happened to all of my colour? Move the Saturation slider to the right until things look good (probably around 20% or so).
NB: If you haven’t calibrated and profiled your monitor, ignore this suggestion because you have no idea what the colours are supposed to look like.
Chromatic Aberration, Fringing, Colour Artifacts
Depending on your camera and lens hardware you may encounter errors with chromatic aberration and/ or with colour fringing – especially around small branches backlit by the sky. In a previous paragraph I mentioned the order of operations and dealing with raw vs. rendered files. Chromatic Aberration is dependent mostly on the lens you’ve used but also on the camera sensor. Capture One Pro does have a Chromatic Aberration reduction feature when dealing with raw files, and the lens data is embedded in the raw file. If you make a panorama in a third party program and import the rendered .tif file into Capture One Pro, often the lens metadata is not carried over and the lens will be set to Generic. You can ask Capture One Pro to analyze the Chromatic Aberration, but you’ll have to select the lens manually from a list.. Capture One Pro has a Purple Fringing slider, but for my equipment I find it inadequate. Fortunately Affinity Photo has a very good Defringe filter. In the image below (300% zoom), on the left is the image in Capture One Pro (Purple Defringe set to 100) and the image on the right has been exported to and defringed using Affinity Photo:
Capture One Pro does have a very good Advanced Colour Editor, and if after defringing you still have residual colours you can often use the Colour Editor to limit their effect.
Another issue you may run into is a colour artifact created by solar flare. This can happen when shooting with the sun in the background and is again hardware dependent.
This is a 14-image stitch made in the rainforest outside of Tofino, BC. As can be clearly seen, there’s a band of magenta and green stretching across the frame. Note that these colour bands were not present in any of the 14 base images; they’re an artifact of the panorama creation. You could try a different panorama stitching software to see if it yields the same effect, but for me, the best way to deal with this issue is to convert the image to grayscale (black and white) as it’s then possible to attenuate the individual colours. For a post on working with Capture One Pro in black and white: Using Capture One Pro in Black and White. Truth be told, converting to black and white can in many cases improve the image – especially for old trees like these.
Here’s a 6-image stitch from John Dean Provincial Park: The lens flare here is pretty much impossible to deal with effectively so you have to decide if the image is worth keeping:
Image Cropping
We’re almost done here, although each one of these issues could easily be expanded upon. Because of the nature of most panorama photography (rotating a camera around one point), when the images are stitched together there will be dark areas outside the frame, like this:
Depending on the panorama software you have, it may automatically crop out these edges for you, it may offer the option to do so, or it may simply leave it to you. There are two ways to deal with these dark areas. One is to open your Crop tool, set the aspect ratio to unconstrained, and crop these black areas out, like this (NB: This image has also been rotated slightly):
Since the edges are black, and since rainforests tend to be dark, it can sometimes be a challenge to find out where the edge of the image actually is. To get around that, I set the Exposure slider to +4EV to make the image as bright as possible before moving the crop boundaries. After setting the crop boundaries I reset Exposure back to wherever it was.
If you have a pixel editor like Affinity Photo, you can use inpainting (content-aware fill) to fill in the black areas with new pixels. In my experience this can be more or less effective, depending on the complexity of what you’re trying to infill. It does a decent job on sky/clouds, but forests, not so much. What I do is to first create a new blank layer and add a transparent rectangle that shows what the crop boundaries would be, like this:
After selecting the black areas around the outside edges of the frame and applying Inpainting, we get here:
One might consider expanding the crop area of this image a little bit, but there are obvious errors, especially at the top and bottom on the right side. I leave it to you to decide.
Keystoning
Okay, one last tip regarding keystoning (no, not the Charlie Chaplin version). Keystoning generally refers to the way tall buildings (and trees, for example) narrow toward the top when we look up at them. Depending on the software you’re using, you can vary the pitch of the image (tilting the top forward or backward) to compensate for that. Buildings tend to have straight lines to work with (Frank Gehry excepted) so keystoning correction is a matter of aligning the guidelines with the edges of the building, window frames, etc. With forests it’s not so simple because trees don’t all grow straight and parallel with each other. Trees grow toward light.
I use Capture One Pro to do most of my post-processing work and the Keystoning tool works very well. Most vertical panoramas of trees will be using a cylindrical projection, which means your image will look something like this:
You can try to use trees on the edges as alignment lines, but as I mentioned, trees aren’t often straight. If shooting with a wide-angle lens, these tilts will be exaggerated. So… you see how the black edges flare out at the bottom of the frame of this two-image stitch? Use those edges for your Keystone correction, like this:
You can see how the Keystoning alignment lines are lined up with the flared edges. When you click Apply, Capture One Pro will use these lines to adjust the pitch and yaw of the image. Other software that offers keystone correction works similarly. Now, the default correction is 100%, but in this case it seemed a bit too much so I toned down the keystone correction to about 75%. After rotating the image slightly, using the Crop tool to crop out the black lines and pushing the sliders around a bit, we end up here:
Final Thoughts
That’s pretty much it for tutorials and tips. I have two final thoughts to leave you with. First, remember that sometimes less is more. Ancient rainforest trees are stately and magnificent, but they show their age and they’re not often ‘pretty’. While I certainly advocate capturing as much of the scene as possible while in the field (you only get one chance to capture a specific scene), especially when making vertical panoramas, it’s better to crop in on the final image.
This is a 21-image stitch of an old cedar tree in the Pacific Rim National Park Reserve. The image on the left shows the entire scene as captured. Aside from the staircase on the left, you can clearly see that the top of the frame is just a mass of branches and has no real aesthetic value in terms of the photograph. The image on the right removes both the staircase and most of the branches.
And finally, just an observation: We’ve been privileged to be around some truly massive trees, each one unique and special in their own right. However, for some reason, particularly with panoramas, the images seem to make the trees smaller than they are. Wide-angle lenses have an enlarged field of view and that’s certainly part of it, but even with a normal lens the effect is still there. I don’t have a good explanation for it and it is what it is, but here’s an example:
This is a four-image stitch of the tree colloquially known as the San Juan Spruce. It has lost its top and so isn’t the largest sitka spruce in BC, but from this image it’s hard to tell that this tree is over 3.7m (12.2 ft) in diameter…
…unless of course one has a beautiful model to stand in for comparison.
Made it this far? Congratulations! If you have any questions or comments, please feel free to send us a message through our contact form and/or to leave a comment! Now go out and make some pictures!!
Hugs,
M&M
P.S. One last, last thing is that one way to improve dynamic range (and other exposure issues) is to shoot bracketed exposures when making the base images for your panorama. You will need at least a sturdy tripod to do this. Once you have the images, depending on your panorama software you may have to create HDR images out of each exposure bracket and then stitch those composites into a panorama, or you may be able to simply combine all of the base images in the stitching software directly. There’s more to this, but if you have any questions please feel free to post a comment or to fill out our contact form.