If you have already read some my instructable here, you know I tend to go for the simplest way to reach an objective. This usually means making affordable and simple projects, but of course I always keep an eye on the results quality.
My passion for immersive photography has born some years ago on the same wavelength, so after a few researches about nodal point, stitching techniques and gears, I've decided to build my own wood panohead and I started the journey which led to an impressive huge 360° images gallery.
Now you have the chance to travel through these steps together. Some of them could be boring, and probably you already have a good experience with many of the described concepts, so I'll try to be more concise and less theoretical as possible.
I want to highlight that this instructable is not supposed to be a personal gallery of my best work, but my ego didn't let me choose ordinary photos ;-)
Step 1: A little bit of Theory
I'm sorry I can't skip the theory step but I won't be mathematical ;-)
The concept behind the spherical panoramas is nothing complicated, this technique is meant to reproduce in virtual reality (VR) the first person view taken in any direction at a certain instant. In simple words you have to take a lot of pictures all around, at full 360° on the horizon and also up (zenith point) and down (nadir point), and merge them into a single image.
Problems appear when you try to merge the pictures together, indeed they have to be warped to match each other. A software could do that easily, but to do that it has to recognize the identical details in different images to overlap them. These identical details are named "control points" (CP). The more you overlap sequential images, the more CP you'll have. To be fair the number of CP is not as essential as their "quality". Years ago I created this panorama of Milan Gallery adding manually 5 CP for each couple of the 16 picture, and the final equirectangular image came out very precise. In case you wondered what equirectangular means, this projection type is the most used way in VR to display a sphere on a plane, and if you'll manage to load it in a panorama viewer it will be certainly supported.
Stitching a set of photos in a uniform single equirectangular image is totally a mathematical issue, and it has been possible thanks to the efforts of Prof. Dr. H. Dersch (who created Panorama Tools) and Andrew Mihal (who created Enblend and Enfuse tools).
If you want to study in depth the subject there are a lot of forums managed by skilled people where you can find really interesting discussions about this process, the most popular are panotools.org and www.kolor.com.
Step 2: The Parallax
If you change your camera position between a shot and the following one, especially if you have objects at various distances, the alignment of these objects will be different, and you'll see two different scenes. To understand this concept simply close an eye and move the head an inch laterally, try to notice what changes in your field of view between the two positions.
If the software will run into these little differences between images, it will have two types of difficulties: an harder work will be needed to find exact identical control points between images it will be much more difficult to build a perfect transition between an image and the adjacent one The stitching programs are very powerful and they'll try to merge all the pictures in a 360° panorama at any cost. This means that in case of wrong alignment you'll obtain an incomprehensible image or an image with a lot of overlapping inaccuracies. I will teach you to avoid both of these situations, and to rectify the few stitching errors that could eventually remain.
Step 3: The Lens
It will surprise you but there is not too much to tell about the lens. The wider the focal distance is, the fewer shots will be needed, but also you'll obtain a smaller resolution on the final panorama. With fish-eye lenses you'll need probably 6 or 8 pictures in total, so the process will be rather easy. I use an 11-18 wide-angle zoom, which lets me obtain a fully 360° panorama from 16-20 pictures at shorter focal length (with an 1.6 crop of the camera).
You can also use a 50mm or a telephoto lens, in that case you'll create an impressive gigapixel panorama. The principle is the same, but you'll need some specific equipment and probably it wouldn't be worth to take a full 360°, because usually details on the sky and on the ground are less interesting.
Step 4: The Leveling
The stitching software usually decides how to align the equirectangular picture on the basis of the first image of the full set. You can certainly change the alignment later, but you can easily get wrong without a vertical line on the scene, or a clear horizon.
A simple and effective solution is to keep a small bubble level on the flash slide to keep the first shot perfectly horizontal, and let the program straighten all the other images consequently.
The two examples show this typical situation: the few subjects that help to straighten the image are trees and the mist, but they're not precise.
Step 5: The No-Parallax Point
The virtual point of your camera lens which you have to keep very still between the shots is the area in which all the light rays will converge. This point is named "projection centre" or "no-parallax point" (NPP). Notice that this is very different from the focal point, which is actually on the plane of the sensor (or the film).
There are many debates about NPP location and a lot of confusion with "nodal point", read the article by Rik Littlefield to take a deeper dive into the issue.
Anyway you can find NPP making certain of the exact alignment of near and distant objects as you see in the animation. For wide angle lenses NPP is very close to the external glass of the lens, for this reason classical tripod heads are ineffective in making precise panoramas, indeed they make the camera rotate around its body, and the NPP will inevitably change position. Also the choice to utilize a tripod mount ring will barely improve the situation, usually tripod mount rings for short lenses don't exist, and however they wouldn't correct the vertical shift.
Step 6: The Panorama Heads
The panorama heads are meant to make the camera rotate around the NPP of the lens, in both horizontal and vertical planes.
It goes without saying that changing lens will change distance between NPP and tripod joint on the camera body. So you have to build your panorama head referring to a certain lens, or adjust the parameters of any universal pano-head when changing lens.
To build a simple panorama head you can refer to my instructable "Light and Cheap Wood Panohead", athough it is not a real step-by-step tutorial because I had already built it at that time.
Step 7: The cheap alternative
You have to be very accurate to go thought the cheapest way... and expect a long post-processing back at home. Anyway that is not unfeasible, indeed about 80% of my panorama are taken with this method. I'm talking about using only a cord and a weight to build a plumb line. Most of the time this is all you need to make an impressive 360° panorama, along with a wide-angle lens of course.
You'll have many benefits using the plumb line, in addition to the money saving:
you can easily keep it in your shirt pocket and mount it in less than a minute you can use it with any lens or camera you can build it very quickly in place if you forget it at home people will not have time to get scared, seeing you mounting the gear, and to run away you can move away if you need, and come back in position with no consequences, e.g. taking a panorama in the middle of the street you can use it in museums or other places where tripod is forbidden, altough taking pictures without a flash will give you a hard time
Step 8: Cheap way downsides
Of course you cannot have everything at the same time, so the cheap method also has a lot of disadvantages:
you obviously will not able to shot gigapixel images, you can use a lens not longer than 20mm you need a lot of light to shot handheld, so you can't use it in low light conditions, nevertheless you can sometimes use the flash you'll surely have big alignment errors, so it works better on open spaces, as a square, without any subject close by you'll be susceptible to strong wind which will make your plumb line an oblique useless plumb line, so you can use a more heavy (and cool) plumb as in picture people who unfortunately hadn't time to run away will look at you as an insane man you are going to waste some minutes explaining to the bravest what you're doing, but they won't understand anyway sometimes you'll stumble on it and it'll take about 20 minutes to untangle the knots, in that case the cool new plumb becomes very useful
Step 9: Overlooking the scene
To make an amazing 360° panorama it's extremely important to find the proper place from where you'll shot all the pictures. Maybe I should say spot rather than place, because even moving the camera a little further will sometimes completely change the view. Write down each of the next statements:
an higher place will let you show much more interesting subjects and avoid to waste half panorama photographing the ground if the ground has a geometric symmetrical motif you can make the image more interesting choosing its central point if you can, try to stay in a shady place, because sun will otherwise be in at least one of the pictures and will create bothersome reflections, with different artifacts in adjacent photos try not to stay too close to any subject, like utility poles, traffic lights, trees, and regular people passing while you're shooting don't forget to pay attention to the electricity cables over the streets, because you will need to check they're aligned in the final panorama audacious locations like the edge of a cliff, the top of a stone, the center of a traffic congested square, the bow of a ship, would give a unique feel to the astonished visitor, but I'm not suggesting you to try to kill yourself! also less dangerous unique spots should make the panorama more interesting, like the center of a drain on the street, a chair in a church, a motorcycle seat, anything that isn't a boring empty space The first panorama shown has been taken from an overhang rock in the Corse Calanches, the second one has been shot from the top of the ancient city walls in Mdina. Both are raised spots.
Step 10: Planning your shots
Found your spot, you can take the pictures. To summarize, if you overlap about 1/3 of a picture with 1/3 of the following one, the program usually works fine. But this evaluation is not absolute, and you have to decide how much to overlap the shots each time, because many factors enters in calculation, as the following ones:
the method you use to take the pictures: if you're using an automatic pano-head you can easily avoid blind areas, so you can reduce overlapping to 10-15% the type of subject: if there are many uniform areas (like sky, fog) or changeable surfaces (like water, flags, animal herds, people crowd, fast clouds) or also recurring elements (like in four cardinal points of Milan Gallery) you'll need a wider overlapping to give the program more CP and maybe more chances to avoid half-body or double-head humans (!!) the variability of light conditions and of the colours: the more you overlap the pictures, the wider the graduated shading between them will be (fast changing cloudy weather, a disco-party with coloured lights...) The same is for both the horizontal and vertical overlapping, but the nadir needs a specific attention.
Step 11: Using Flash
You can usually light up your scene with a flash, which is undoubtedly better than abstain from shooting the panorama. Anyway, if you use a a flash attached to the camera you'll obtain different shadows in adjacent shots, unless you manage to make the flash pivot on the NPP (no-parallax point), unfortunately that's everything but simple... The panorama from the inside of a pub on the Italian Riviera has been taken handheld with the camera incorporated flash, but I had to make a lot of postprocessing to fix the bad shadows overlapping, and the program worked hard to find good control points.
Another way could be lighting up the full scene with an indirect light from one or more fixed flashes, that would work very good. I've built a DIY flash diffuser to spread light all around the room, you can keep it under the tripod paying attention to the tripod legs shadows.
Step 12: How to become a ghost
Obviously you'll surely expect to find the tripod or the photographer feet at the bottom side of any 360°panorama... and indeed you'll find it in many cases. But any professional and meticulous photographer will abhor that! So it's essential to find a way to hide these unrelated elements. You have three possibilities: you can replace them with a more professional element, like a circle or a sphere with your contact data:you've to elaborate the final image to do this, and we'll see that in last steps you can recreate the ground or the pavement with some postprocessing work: we'll see that too you can try to empty the area under your camera when you shoot the nadir: in this case you have to take more shots (maybe 4 instead of one if your're using a wide-angle lens like mine) moving away your feet or tilting the tripod so that neither shoes or tripod legs will appear in images
The last technique is obviously the one that lets you stitch the images faster, but it's not always possible.
I bet that if you are a diligent student you have already noticed that ghosts have no shadow, hiding your own shadow needs the same trickery as the nadir cleaning, look the following animated gif to see how to take two different shots so that the software hide the most part of the shadow, then you have to hide the remaining part by hand.
Step 13: Setting up the images
Before starting to analize stitching programs, I want to make sure that the way to obtain maximum quality from your camera is clear to you. Since shooting and postprocessing spherical panoramas takes a long time, taking many pictures and trashing the worst ones is surely worth it. I think it's a good choice to shoot a bracketing set of RAW images. This will occupy a huge space on your memory card, but they're much cheaper now, if you don't take part to the transfer speed race. You can buy good 64Gb Industrial Transcend CF at a very cheap price, and although they're not specifically made for photographers, they work very well and are fast enough for most situations.
In this case I shot a bracketing sequence of pictures, but then I decided to keep only the lighter set, and discard the other two. Of course feel free to follow my "exposure bracketing for spectacular panoramas" instructable to add appeal to your panorama.
I also tried to use polarized filter this time but as you can see, it creates an unsuitable variation of the sky tonality, due to the different angle between the filter and light rays in adjacent photos, even though I tried to rotate it properly for each shot. Result: don't use polarized filter to make panoramas.
Lightroom is a professional but very intuitive program to develop RAW images to obtain the maximum color quality and detail. Here you can see the three main steps to use it:
importing raw files into library developing them adjusting color, exposure, contrast, etc. exporting them as jpg files (your stitching program will need less time to transform jpeg files, but in case you don't care about time, you should save them as 16bit tiff)
Step 14: Ready to be stitched
Now you have a full set of images (I have two, because I wanted to explain you the process through two different stitching programs), ready to be merged together in a full 360° spherical panorama.
I prefer not to examine in depth the process of stitching pictures using a program rather than another. This is because I don't want to limit the choice to a single software, and you can find hundreds of guides about this around the web.
So I decided to explain you just the basics of two great programs: Hugin and Autopano. Autopano is a more recent and professional powerful program, but Hugin lets you understand the stitching technique in a better way, and you can set many parameters of the process, plus it's free for commercial use, and it has a long story in his past.
You'll have to try them a little before deciding which one is the best choice for you. The stitching method may vary depending on personal liking, I believe that merging pictures and obtaining the equirectangular final image has to be fun other than efficient, this way you can feel inspired and create wonderful pieces. The fact that these two panoramas are not masterpieces will prove my thesis: I quickly made them just to illustrate my tutorial, but I wasn’t very inspired ;-)
Step 15: Hugin
It should be useful for you to look at the images and follow the process from the loading of the single pictures to the rendering of the final equirectangular image. Don’t forget to read the notes on my pictures so you will be able to understand how this program works.
Hugin is not an intuitive program, this is because it leaves you a large choice for most calculation parameters. I’m not going to explain you advanced techniques here, you will be able to enhance your understanding with time. Original settings should be enough for beginners.
Step 16: Autopano
Autopano is a very powerful software so you'll hear your computer working hard (and fast) as it stitches pictures and renders your panorama. Autopano can even merge thousands of images in a single gigapixel panorama. Anyway it should handle your two dozen pictures without efforts. If you have taken every photo of the panorama properly, you will get a very good result in a minute or two. In case you took pictures moving your camera with your own hands like I did (there was an incredible wind that pushed my plumb-line away...), you'll have to do some work, adding and deleting CP between some images, and you'll have to do some postprocessing to remove little inaccuracies.
Step 17: Fixing the nadir
As I anticipated before, if your efforts to become a ghost didn't work, you need to edit the equirectangular image to remove your feet. Because the equirectangular picture is a planar projection of the sphere, it reminds the world map and in that exact way it shows the poles stretched all along the length of the map. The nadir is nothing more than the south pole, and your feet will be stretched as BigFoot feet.
If you want to go for the easy way you can simply use Photoshop to fill the empty bottom area of the equirectangular projection with the specular image of the same view, and your can add you website address too. This is quite professional, because you cover the less interesting part of the panorama with the logo, and people are usually happy to find something more interesting there. Follow the tips on the images to better understand the entire process.
Step 18: The Specular Dome
The solution of the specular dome can be very effective, as you can see in this other example. Here I added a transparency at the sphere borders to reveal the underneath tiles. Always remember to add a shadow over the overlapped rectangle, and a faded layer of a black or bright tint, depending on the effect that you want to give. As usual, my suggestion is to test different settings.
But if you're a real obsessive perfectionist, you need to convert the pole to an ordinary view, modify it (recreating the pavement, copying the grass, deleting your shadow, etc.) and converting it back to equirectangular.
To do that without specific filters in Photoshop, you can flip the image vertically and apply the Polar Coordinates filter (in the Distort group) with Rectangular to Polar selection, although this will deform your image a little. However, you can clone the grass or dirt and apply the same filter with Polar to Rectangular selected, and of course mirror the image again.
Step 19: Perfect geometric motif
If you have a geometric pavement, a bench, or a sidewalk with parking lines, you need to see perfectly parallel lines. This way it won’t be hard for you to clone the motif and replace your feet. This is much more difficult to obtain.
The most intuitive tool to extract a parallel view of any part of your spherical panorama is PTEditor. This is indeed a Java application included in Panorama Tools, and you need JRE to run it (in my case it worked with "jre-7u5-windows-i586" on my Win7 64bit, but many people have incurred in problems, read here for alternative methods...).
This application is very easy to use, and it's made exactly for that purpose:
If your panorama is wider than 8000 px you have to make a smaller copy to load into PTEditor, in my case it crashes when I use bigger images, but keep your original file, you'll need it! After opening your reduced panorama into PTEditor you can turn around the view dragging on the image, and zooming in and out with shft or ctrl+left click. When you have the exact frame, including the zone you want to modify and an additional sample pavement around that, you can Edit -> Extract Partial View. You have a tiff image now (maybe you have to rename it if you opened a jpg in PTEditor) and you can retouch it in Photoshop. After you've fixed the floor/ground/pavement/chair you can save the tiff and reload it into PTEditor with Edit -> Insert Saved View, then save the panorama and close PTEditor. Now, if you had to reduce the panorama, you should merge the new pavement area on the original file, simply overlay the new equirectangular image (enlarging it) on the original one, and finally hide everything with a mask except the fixed region.
Step 20: The result
At the end, you should have a precise, good quality equirectangular picture to upload on a panorama site, that will show it to world communities using a good quality proper viewer like krpano. This great flash 5 engine lets you looking at the spherical panorama assigning different projection types. Feel free to experiment Fisheye, Architectural, Little Planet view, and other visualization types, right-clicking on the screen of the Milan Gallery panorama, the same as many other examples included in krpano galleries.
Step 21: The alternative views
If you hadn't already noticed, you have achieved much more than just a spherical view of the scene, you now have the opportunity to obtain some really unique framing and perspectives from your equirectangular file, something that usually belongs to professional world photography. I'm referring particularly to architectural views with parallel vertical and horizontal lines and fish-eye pictures that are otherwise obtained using expensive lenses, and you can also have a peculiar view of the scene from a virtually high spot.
To reach these results you can employ Hugin again, loading the equirectangular file as a singular image and assigning the right properties (equirectangular projection type, 360° horizontal field of view). In the preview, you can change the resulting projection type into fish-eye, architectural, Mercator, etc. and drag the image to change the view direction.
Attached pictures are in this order: a mini-world panorama taken on the biggest lake of Italy, a view of a castle on Dolomites, a stereographic view of the Pantheon of Paris, and a similar shot taken in an ancient building in Venice.
I hope you enjoyed this short/long, essential/exhaustive, boring/interesting, useless/ultimate guide about this awesome photographic technique. Write me or leave me a feedback if you wish.