Multi Image Compositions -- Panoramas

One of my goals in Photography is to create the largest, most detailed rendering of a scene. I am forever seeking the best cameras and lenses to create the detail I seek. My definition of success is when an observer of my work reacts with "Wow ! just look at those details, what clarity". Of course its not just the camera and the lenses that contribute, everything from subject matter to post processing and printing align to carry the details to the final product, be it web display or print. 

Having enough detail at the print or display stage is critical. Printing large on printers that support 250 to 300 dpi (dots per inch) requires large image files. A 20x30 in. image optimized to print on a 300 dpi printer requires 54 MegaPixels of information, similarly a 24x36 in. print requires 77.7 MegaPixels. While it can be subjective as whether the human eye can perceive the difference between a 150 or 300 dpi image, more information is always better. The more MegaPixels, if recorded appropriately, will always produce a crisper image. 

To achieve such detailed rendering given the limited sensors sized and lenses of current model cameras lead me to combining multiple images into a larger, more detailed, rendering of a scene. Often this is referred to as "shooting a Panorama" however panoramas are otherwise defined as a picture that has a width to height (or aspect) ratio of 2:1 or larger. So while multiple images may be (and often are) used to create panoramas, not all panoramas require multiple images. Similarly not all multi image scenes are panoramic in display.

Multi image compositions are the result of taking multiple overlapping images of a scene and assembling them in post processing to create a image with more detail than would be possible with a single picture. Given the technology available for both taking and post processing images it is now possible to generate Gigapixel images with stunning resolution using simple point and shoot cameras. Given the ability and a camera the only additional requirements are a bit of patience, some software and a reasonable computer for post processing.

The interesting part of multi image compositions is that it is easy to go too far in recording details. Without much effort is it possible to create images that exceed the capability for display. My recommendation is to start backward if you will -- decide what you want your output to be, then plan how much information you need to achieve that output. I find that "working backwards" helps me when I compose in the field and brings me more success at achieving my final product.

Most cameras today record images at a fixed 3:2 aspect ratio. In single capture imaging you optimize the focal length and your distance to appropriately frame the subject. If the 3:2 ratio does not emphasize your subject as desired, you can crop the image in post processing.

When capturing a scene by stitching multiple images a bit of forethought at the time of image capture can aid in the overall success of the effort. Unlike the single shot capture you have several choices of focal lengths to choose from in recording a scene. As an example, if you can capture the entire subject with a 50mm lens using a single shot you could capture -- ignoring necessary overlap -- the same subject with four shots using a 100mm lens or 16 shots with a 200mm lens. Capturing the scene with the 100m lens will give you 4 times the pixels with twice the magnification. Similarly at 200mm you would end with 16 times the pixels at 4 times the magnification. With each increase in focal length and number of panels there is a corresponding increase in number of images and thus resolution. The downside is the increased workload; one to four then eight images for our hypothetical scenes. Knowing the desired aspect ratio and some idea of the desired print size can help you choose the resolution and number of panels to combine. Simply put, by planning ahead you can reduce the overall workload. While all this "prior planning" may sound extreme much of this can be done by understanding your equipment and building up some simple rules and concepts.

To illustrate with another example, my workhorse camera is a Canon 1ds2 sporting 16.7 Megapixels. The sensor records a capture with 3328 x 4992 pixels which will produce an 11x16 print at 300 dpi. From this I know that if I use my camera in landscape orientation the most I will be able to print (in the vertical) is 11 inches. If I use the camera in portrait mode I will get 16 inches. 

The image below captures the Baltimore Inner Harbor from Federal Hill during a winter sunset. When taking this image I plan ahead -- I know in the end that I will be making a 2:1 or 3:1 aspect ratio final print. I also know that at 3:1 my maximum print size would be 15 inches in the vertical -- giving a 45 inch wide print.

Knowing this I realize that I can either capture the shot using two rows of images with the camera in the landscape orientation or a single row of overlapping images in the vertical orientation. By going portrait I get enough vertical resolution out of each shot and I simply need to stitch several shorts horizontally to get my desired width.

The constituent shots I used to assemble this image are shown below. Note that in this case there was a good amount of cropping in the vertical dimension which means my focal length was not optimized for the scene. I shot this with a 50mm lens and I would have been better served using the 85mm lens instead.

Rule #1 -- Use Manual Mode

Matching exposure is critical across the scene. The only effective way to record the scene properly is to use manual exposure mode on your camera where you explicitly control the aperture and shutter speed. If you are unsure how to use manual mode you will need dig into your camera's manual. One trick is to use Av or aperture mode to obtain the right settings for the scene. In aperture mode you select the appropriate aperture -- often f5.6 to f11 and the camera will choose the correct shutter speed. I usually use this method, scanning the entire scene to find the brightest object or part of the scene and then remember the settings when switching to manual mode. 

Rule #2 -- Overlap the Images

To get good stitching I regularly use 25% overlap. In many cases I will do 75% overlap which effectively gives me two complete sets of images recording the same scene. This has saved me on several occasion when one of the photos in a set was blurry or did not focus well.

Rule #3 -- Use a Tripod and Remote Shutter Release

The whole reason you are recording the scene with multiple captures is to get the most detailed result. While hand held multi capture images are certainly possible, use of a tripod and remote shutter is an necessary part of the process.