Producing PanoramasTom Arah shows you how to create successful panoramas and looks to the future with immersive and object-based imaging.
A single shot completely fails to capture the wide field of view of a sweeping panorama.
Every year hundreds of millions of photographs are taken to make a permanent record of a temporary view. And every single time they fail to do justice to the actual experience. Part of the problem is that the photograph is flat and static whereas our vision is full of depth and movement. Less obvious but just as fundamental a limitation is the fact that the ubiquitous 35mm print is just a small 3x4 ratio rectangular section of what we actually see. When you compare looking at the photograph of a view to the actual visual experience of the sweeping vista that inspired it, the point really hits home. It's a cliché but a photo just isn't the same as being there.
Of course the obvious first step in improving the situation is to take a panorama shot, but what's the best way? This might not seem an issue these days as every APS camera sold now offers a dedicated panorama mode. In fact these aren't true panoramas at all as the system simply crops the top and bottom of the picture so that you actually end up seeing less rather than more. To really expand your field of view (FOV) with traditional photography you need a dedicated panorama camera. Over the years every conceivable way of doing this has been attempted, most involving rotating the whole camera or just the lens and effectively painting the final image onto the film. The problem with this is that the cameras are inconvenient, expensive and incapable of producing normal shots.
Taking the panorama directly in the camera isn't an option for the average user, but there is an alternative solution tailor-made for the new generation of digital cameras: produce it on the computer. To create an APS-style faux-panorama effect, for example, you can simply interactively crop to the central band of interest. More importantly, by joining multiple photographs together you can produce true wide FOV panoramas. Even better, by stitching photographs, you are effectively increasing your final image's resolution so that you kill two birds with one stone: simultaneously improving image content and quality.
Taking The Shots
The idea of stitching photographs together on your computer sounds simple, but the practice is actually pretty demanding - starting from the moment you take your component shots. Essentially you want to capture a series of images around a single point of rotation, so a tripod is ideal. Fortunately it isn't essential for panoramas of less than 180° so long as you keep the camera as level as possible and use the optical viewfinder rather than the screen viewfinder to keep the movement as you rotate to a minimum. You want a good area of overlap on each shot, preferably a third or more, and you also want to try and avoid cutting up strong features especially if they are in the foreground. It's much better to take too many shots rather than too few.
Taking multiple shots also has the major advantage that it helps cut down distortion. If you think about looking at a fence, at the centre of your vision the slats of the fence produce regular rectangles, but on either side perspective turns these rectangles into trapezia with each fence post progressively smaller into the distance. Trying to match these from one photograph to another would be impossible, so you want to be able to work with just the head-on central sections especially if your camera has a wide-angle lens which exaggerates the effect. If your camera doesn't have a fixed lens it's also better to, if necessary step back, and zoom in and take more images more flat on.
Closely spaced photographs also help cut down on the widely varying results produced by automatic exposure cameras. Auto exposure compensates for different lighting conditions by adjusting the tonal range between shadows and highlights. This produces the best results for each separate picture but means that the colours of one shot can be completely different to the next. This problem can be reduced by shooting when it's cloudy but still bright, or at midday when the lighting is even. Much the best solution, if your camera offers the facility, is to lock your exposure based on the shot with the most typical lighting conditions. You'll end up with a slight loss of detail in both shadows and highlights, but this is more than made up for by more even exposures and colours.
Now that we've got the best possible material it's time to see what computer manipulation can do. Photoshop doesn't offer any dedicated panorama-based features but that needn't stop us. To begin with we need to set up a canvas wide enough to take all of our photos less their overlaps but with a bit extra to spare on all sides. Then we need to load each shot as a separate layer which is most easily done by opening all images and then cutting and pasting them onto our canvas. The layers can then be selected in turn by Ctrl-clicking on them with the Move tool and positioned over each other so that the major features line up as best as possible. To make the process easier you can make each layer semi-transparent.
Manual stitching can be a complex business
Next we have to sort out the seams where the images overlap. Essentially we're trying to blend from one picture to another so we can start with a gradual dissolve between the two. To achieve this you need to add a layer mask and then add a linear gradient from black to white to match the overlap. For featureless areas such as clouds, sea and grass this blending works well but it's hopeless for the important features in the panorama which just end up blurred. Here you have to decide to pull out the feature from either one photo or the other. By swapping between solid black and solid white you can swap between revealing one photo or the other while painting on the layer mask.
Working in this way you can achieve a seam that is a best-fit compromise between the two overlaps but the results will usually need more interactive fine-tuning. To try and sort out distortion and exposure problems, you'll first have to select the visible area of each layer, by Ctrl-clicking on the layer name. You can then use the Free Transform option to distort it toward the desired shape or add a ready-masked adjustment layer to try and massage the tonal range back into line. When you're as happy as you're going to be, save a flattened copy of the entire image and, where necessary, go in to tidy up with the clone and smudge tools. The system works reasonably well, especially for distant vistas with well-spaced features and even exposures, but for more close-up and busy work prepare to be frustrated. By their nature photographs taken at different angles just don't line up.
Obviously we could do with a bit more help, and a number of packages now offer dedicated panorama features. In most cases though this boils down to nothing more than automatically creating a suitably wide canvas and evenly spacing the component images. Recently though my eyes were opened by the new stitching feature in PhotoSuite III. This worked like magic, producing much better results than you can hope to achieve manually and in a fraction of the time. The feature was based on ImageSynthesis technology produced by Enroute so I visited their web site (www.enroute.com) and tracked down their own dedicated package QuickStitch 2.
Enroute QuickStitch can stitch both horizontally and vertically - and without any pain.
The program is simplicity itself. You simply select your component images and drop them in the right order onto a grid. You then choose between "panhead" for wide FOV panoramas or "perspective" for narrow FOV panoramas where you want lines to remain rectilinear, and set your desired output size ranging from "small", 320 x 240, to "jumbo" at over 6 mega-pixels. You then click on the Stitch button and after a few seconds your panorama appears in all its glory. The real beauty of the system is its latitude. QuickStitch doesn't need information on camera settings or accurate image alignment or perfectly even exposures. It is also unique in that it enables both vertical and horizontal stitching simultaneously, up to six images wide and four high, which is ideal if you're primarily looking to boost an aging camera's image resolution/quality.
The program is simple in practice but clearly there's a lot more happening behind the scenes. Enroute doesn't give much away about how its ImageSynthesis technology works, but it's possible to make some intelligent guesses. The system works best with "featureful" images so it's not primarily pixel-blending as we were in Photoshop. The program also prefers images with a 50% overlap on either side so that in the middle of the panorama it's working with two references for each point. Essentially I imagine the system as working rather like the brain taking the raw information from two sources, pulling out the content features from each together with information on angles and so depth and then manipulating and merging these back into a single, "mathematically correct" unified image. As such it's not surprising to find that the technology is largely derived from work by Enroute's CEO, Leo Blume, in satellite mapping - and that the company has a development team in Moscow.
QuickTime / Immersive
Another advantage of the full QuickStitch 2 software is that if you produce a panhead wide-angle panorama you can save it as a QuickTime Virtual Reality (QTVR) movie. The dialog for this provides a number of options for controlling quality, compression method and so on, but also allows you to set the resulting movie's window size and its field of view which determines what the user will see when they open the file.
Saving to QTVR enables the creation of immersive movies.
The difference is amazing. So far the panoramas we have been talking about have all been false to an extent as they are presented as both flat and head on which isn't actually how we experience the scene at all. With a panorama in real life you turn your head to look at the full vista or you stop and focus on a particular element of interest. With the QTVR movie the panorama is curved just as in real life and to see the rest of the view you pan and, if something catches your eye, you zoom in on it. The overall effect is that you aren't just looking at a flat image but are immersed in a realistic scene.
The obvious logical conclusion for QTVR is to produce a full 360° panorama, but irritatingly the basic QuickStitch 2 doesn't stretch this far. To produce fully immersive panoramas like this then you'll have to buy the separate QuickStitch 360 or a similar program such as PhotoVista. To see immersive panoramas in operation (though without the zoom) you can visit www.thegen.com. This is an ambitious shockwave-based site that plans to enable browsers to virtually visit a whole range of cities before deciding whether you want to go there in person. Quite sensibly it's started at the top with the most beautiful city of them all - Edinburgh. 40,000 images have been turned into a range of 360° panoramas showing the views from the Castle, Princes Street and so on.
Immersive panoramas are best experienced via the Web.
The site undoubtedly shows the potential of such virtual tourism and no doubt estate and travel agents across the country will be dreaming of similar schemes. However there's a problem. The 360° panoramas are certainly much more realistic than flat panoramas, but there's still no real depth. With QTVR panoramas you can zoom in but you can't actually move; your viewing position is fixed. Clearly a real tour of a virtual city or house would let you move through it. The requirement is even more urgent for the whole area of online shopping where it's essential that buyers should be able to see a potential purchase from all sides, effectively handling the object as they would in a shop. What's needed is some way of marrying the realism of 2D photography to the viewing flexibility of 3D.
In fact the QTVR format is able to offer this, at least in a limited sense, with its object photography capabilities. The same general image stitching idea is used but this time the images are all taken at precisely defined angles which allows a 360° navigable movie of the object to be produced. This is a big step forward but the drawbacks are equally obvious. To begin with, the creation process is impossible without an object rig with an indexable turntable and swing arm. More importantly, the end results are as fixed in their way as the QTVR panorama. You're not going to be able to create your truly navigable virtual reality house or city view like this.
A far more impressive solution is Metacreations' Canoma technology which joins the idea of image stitching to true 3D modelling. What's amazing about the system is that it not only works but it actually simplifies the whole process of 3D modelling. Essentially you load a photograph and drag 3D primitives, such as cubes and pyramids, onto the image and pin the object's corners and edges to match the image's main features. You then take another image of the same scene taken from any other angle and match the same primitives to it. When you then save your scene to a Metastream or VRML2 format file, Canoma saves the VR wireframes of all objects and uses the photographs to build their surface bitmap textures.
Metacreations Canoma turns photographs into true VR scenes.
The system is beautifully simple and means that you can either produce a QTVR-style object movie with two photographs and five minutes' work or, with more photographs and more effort, you can produce a fully navigable, photo-realistic, multiple-object VR scene. This is serious power, especially when allied with Metacreations scalable MetaStream browser technology, but it's still not perfect. To begin with the primitives offered in Canoma are just too primitive. If you look around you you'll see that real life is far more complex than simple cubes and pyramids and the Canoma approach really depends on flat surfaces for its success. How, for example, could you create a model of a face? More to the point, life is just too short. If you look about your office and think about the amount of work that would be needed to realistically map every screen, table, chair and so on, the effort involved would be horrendous.
Clearly for the future the whole process of stitching and modeling must be made completely automatic. In terms of photo stitching, a number of camera manufacturers are working on just this, putting ImageSynthesis-style technology directly in the camera. This will enable true panoramic modes where a sequence of shots is automatically joined together, effectively lifting the digital camera resolution restriction. The more shots there are and the more they overlap the better so it's not surprising to find stitching technology turning to video. The most exciting work in this area is again being done by Enroute with its video panorama system. Effectively this combines the panorama and object photography capabilities of QTVR by taking feeds from multiple cameras and stitching them into a spherical video. Using the viewer program, users can then zoom in and pan around the live picture effectively acting as directors of their own movie.
Just as promising are the developments for automating the 3D modelling process. Again Metacreations is the company in the vanguard with its MetaFlash technology. This is a modified flash attachment for digital cameras that pulls out depth information and so automatically builds up the necessary 3D wire frame models onto which the photo-based texture maps are then applied. The potential of making 3D modelling "almost as easy as taking a photo" is immense and Intel and Kodak amongst others have jumped on the bandwagon by licensing the technology. The obvious limitations still to overcome are the complexity of matching models taken from different angles and the dependence on a flash which seriously restricts the scope of the system.
Of course the ultimate goal will be to completely integrate the two stitching and modelling technologies to create completely realistic VR environments. The ideal would be for an estate agent, for example, to be able to walk through a house panning a video camera around the rooms to build up a combined visual and 3D scene. The processing power required would be enormous, but it shouldn't be impossible - after all that's exactly what our eyes and brain are doing every second of every day. The difference to our everyday vision would be that the VR scene would provide a truly permanent record that, using 3D goggles or similar viewing equipment, could be re-experienced just as freshly as it was the first time.
More importantly, exactly the same real life experience could be shared and relived by others. Who knows maybe in future you'll even be pleased when your neighbours invite you around to take a look at their holiday pictures.
Hopefully you've found the information you were looking for. For further information please click here.
For free trials and special offers please click the following recommended links:
For further information on the following design applications and subjects please click on the links below:
[3D], [3ds max], [Adobe], [Acrobat], [Cinema 4D], [Corel], [CorelDRAW], [Creative Suite], [Digital Image], [Dreamweaver], [Director], [Fireworks], [Flash], [FreeHand], [FrameMaker], [FrontPage], [GoLive], [Graphic Design], [HTML/CSS], [Illustrator], [InDesign], [Macromedia], [Macromedia Studio], [Microsoft], [NetObjects Fusion], [PageMaker], [Paint Shop Pro], [Painter], [Photo Editing], [PhotoImpact], [Photoshop], [Photoshop Elements], [Publisher], [QuarkXPress], [Web Design]
To continue your search on the designer-info.com site and beyond please use the Google and Amazon search boxes below:
|designer-info.com: independent, informed, intelligent, incisive, in-depth...|
All the work on the site (over 250 reviews, over 100 articles and tutorials) has been written by me, Tom Arah It's also me who maintains the site, answers your emails etc. The site is very popular and from your feedback I know it's a useful resource - but it takes a lot to keep it up.
You can help keep the site running, independent and free by Bookmarking the site (if you don't you might never find it again), telling others about it and by coming back (new content is added every month). Even better you can make a donation eg $5 the typical cost of just one issue of a print magazine or buy anything via Amazon.com or Amazon.co.uk (now or next time you feel like shopping) using these links or the designer-info.com shop - it's a great way of quickly finding the best buys, it costs you nothing and I gain a small but much-appreciated commission.
Thanks very much, Tom Arah
[DTP/Publishing] [Vector Drawing] [Bitmap/Photo] [Web] [3D]
[Articles/Tutorials] [Reviews/Archive] [Shop] [Home/What's New]