A Panoramic Photography Workflow
This post is to share my personal approach and workflow for creating stitched panoramic images. People that practice panoramic photography often develop a set of techniques that work well for them. These are techniques that I have developed based on my own experiences and my own preferences - this is what works for me.
Workflow Part 1: Get That Lens Level!
There are countless tutorials on the internet that discuss the importance of lens alignment when capturing your panoramic scenes. But I don't recall any tutorial that visually illustrates the importance of lens alignment. I explain the importance of lens alignment in the context of your final composition. The bottom line is your lens angle will have a significant impact on your final composition (or final crop). The illustration below is exaggerated to demonstrate several scenarios of how poor lens alignment can cause problems in the overall project.
Fig. A: This example assumes that 5 panels were captured for the panoramic stitch, and assumes the lens is angled upward. Problem #1 is Perspective Distortion: Consider the effect that an upward camera angle has on vertical elements - the lens will bend verticals inward as the verticals rise from bottom-to-top.
The primary problem with perspective distortion in panoramic photography is what happens during the stitching process. It is possible for the stitching software to attempt to correct perspective distortion by warping, bending, repositioning, and cropping each panel in order to make all of the panels align. The 5 blue panels shown in Fig. A at left are exaggerated to show gaps that may occur between the panels if the stitching software is successful in correcting all of the vertical perspective distortion. This impacts your final composition in a big way since you now have to crop away those gaps or clone the gaps from other parts of the image. Problem #2 is Lens Arch: In my opinion, the lens arching as the camera pans is a bigger problem than perspective distortion. I choose to pan my camera from left-to-right. Fig. A shows the arch that will occur if your lens is angled upward. Your stitched image will show evidence of the arch and you're faced with another big problem for your final composition. Again, you either have to crop away all of the empty space around the stitched image or clone all of the blank spaces from other areas of the image.
Fig. A shows two problems that arise from poor lens alignment when cropping your final composition. The yellow box represents the overall canvas size that your stitching software might create. The initial canvas size will be determined by the outer coordinates of the pixels used from each of the individual panels. Now you really only have a few options. 1) Attempt to use the entire canvas by cloning away all of the empty space (the yellow space), or by layering in bits and pieces from your original files. Both can be difficult to achieve. 2) Make a HUGE crop by cropping inside of the stitched pixels. I've seen this done a lot and it usually creates a terrible aspect ratio and sometimes negates the effort of creating a panoramic image in the first place. 3) Use a crop that is somewhere between 1 and 2.
NOTE: Depending on the scope of the project, there will be times when the camera must arch. If that is the case, you just need to plan on capturing more rows of panels to fill in the gaps above and below the arch.
Workflow Part 2: Evaluate the Size of the Scene
Fig. B: Click to enlarge Fig. B: Evaluate your scene to determine how many panels you will need to capture. Most tutorials I've read about panoramic photography suggest 25% - 30% overlap across the panels. In my opinion, that isn't enough. Again, consider your final composition and final crop. One of the phenomenons that occur in the stitching process is that your panels may take on an oval shape. I don't know the science behind this problem but it has something to do with the rotating motion of the lens as it rotates on your tripod (in a cylindrical fashion), and then the stitching software trying to make a flat image from something that was captured "in a cylinder". Fig. B shows two crops that are available depending on how many panels were captured. As before the yellow box represents the overall canvas size and as before, you either have to crop away the empty spaces or clone them in. Each scenario is fairly self-explanatory. I choose the second scenario by overlapping my panels by 50% so that I have more pixels to fill in the gaps between the panels.
NOTE: Fig. B emphasizes each panel in portrait orientation. I choose to capture my panoramas in portrait orientation so that I can gain maximum height across the image overall. This gives me a lot more freedom with composing in-camera and for final crop.
Part 3: Evaluate the Overall Exposure
I consider how the exposure may vary across the entire scene. Depending on your angle to the light, it is possible for one side of the scene to be much lighter than the other side of the scene. It is a matter of personal preference how you expose each of the panels. But you need to be aware how the light changes across the scene so that you can decide if your exposures should vary as you pan the camera. I make this decision depending on the complexity of the light for each scene. Below are two examples to make the point.
Fig. C: In this blue-hour scene the sun was very low on the horizon. The sun is just out of the frame to the left. By evaluating this scene I found that the overall exposure varied by two full stops from left-to-right, and in this particular case I chose to vary my exposures to compensate for the changes in light. Of course it may have been more "realistic" to photograph the scene as it was but I considered that the right side of the scene may have been very dark and may have upset the final product. Right or wrong, it is your choice how to manage this problem. Fig. D illustrates what the scene may have looked like if I had not varied my exposures while panning.
Part 3A: Evaluate the Overall Exposure, continued
To "HDR", or not to "HDR" - that is the question. When it comes to panoramic photography, I approach the in-camera work the same as I would for a single-frame image. Depending on the challenges for a given scene, I will photograph the scene one of two ways: 1) The stitch will use single-exposure panels. Or, 2) The stitch will use panels that are created from multiple-exposure tone-blended images. I consider the techniques I use for HDR photography as an arrow in my quiver. I simply use the tool that I need to complete the vision I have for a given scene. Although I regularly rely on the dynamic range of my raw files for single-exposure images I have found that I can get far superior results (in terms of image quality) by using my HDR techniques if and when the scene demands it. The cityscape image above is a good example of why I might choose the HDR route. The bottom line is that I wanted to see a lot of detail in the buildings, bridges, and reflections in the water but the buildings were mostly back lit given the position of the sun. Although I could have stretched the tones in single-exposure raw files, taking the HDR route gave me much better control over the signal-to-noise ratio. I metered 5 stops of difference between the highlights and shadows in this scene, and that's what each panel received - 5 exposures at +/- 1.0 EV. Fig. F below illustrates the dynamic range that was captured for each panel.
Fig. G: Click to enlarge Most of my panoramas are stitched from single-exposure panels. This is especially easy to achieve on overcast days. The light in Fig. G to the right also varied in overall EV from side-to-side, but only by approximately 2/3 a stop. I knew I could easily adjust the highlights and shadows of each raw file to compensate for the differences in light. So in this case I locked the exposure and simply panned the camera.
Part 4: Determine the Focus Point
I have had cases where I was able to use a locked focus throughout the capture process, and I have had cases where I chose to adjust the focus point throughout the capture process. Using a locked focus point: The cityscape above (Fig. C) is an example where I used a single AF point that was locked through the capture. A good understanding of hyperfocal distance can be very useful in this area. The camera is approximately 1,600 feet from the cluster of buildings in the middle of the frame. I chose to photograph the scene at 24mm. Frankly, at that distance and using 24mm, I could have shot it at f/2.8 and still have sharp results. But in this case I shot at f/11 considering the anticipated sharpness with that lens at 24mm. In either case I knew the hyperfocal distance would carry the entire depth of the scene. So I used a single AF point, locked the focus on the buildings, and locked the focus to manual. If you do this, just be careful not to touch the focus ring throughout the capture or you'll basically have to start all over. I have used a piece of tape placed across the focus ring to prevent accidentally changing the focus in cases like this. Using a variable focus point: I varied the focus throughout the capture process in Fig. G above. I chose to do that because much of the subject was physically closer to the camera than the rest of the subject. In this case I used the full AF array and refocused the lens with each panel. In a few panels I refocused the lens multiple times to balance the number of AF points that were locked onto the subject. As I mentioned before I overlap my panels by as much as 50%. Some stitching software programs use on the sharpest pixels from overlapping frames when merging the panels, as was the case in CS5 at the time this image was stitched together. Assuming CS5 did its' job, Fig. G is comprised of only the sharpest pixels front-to-back, side-to-side, and top-to-bottom.
Part 5: Set your White Balance to MANUAL
Iif you're not familiar with white balance, I suggest reading this article. I've seen a lot panoramic images ruined when shooting on auto white balance. It is possible that if using auto white balance, the camera may pick up a shift in light quality or color signal as the camera pans, and suddenly you have one or more panel that doesn't match the white balance of the other panels. Using manual white balance can reduce the risk of this happening. Fig. C and Fig. G above were both captured using "cloudy" white balance.
Part 6: Capture the Scene
Part 7: Post Processing
Post processing is probably the biggest area where people's processes vary. Since I shoot in raw, I intend to maximize the quality of the stitch by processing each raw file prior to stitching. Above I mentioned single-exposure projects versus HDR projects. My raw processing for each scenario looks something like this:
A) Single-exposure panels: Lens corrections - white balance - highlights - shadows - contrast - clarity - hue/saturation/lightness of color - noise reduction - output sharpening. Lightroom makes it easy to copy and paste the adjustment settings to multiple files. So I only manually adjust one panel and copy and paste the settings to the other panels.
B) Multiple-exposure tone-blended panels: Lens corrections - white balance - clarity - noise reduction - output sharpening. Since each panel will be a blend of multiple exposures, I don't adjust highlight, shadows, contrast, or colors before stitching. All of that information comes from the multiple exposures that are about to be blended for each panel.
Stitching the Panels: There are a lot of fine stitching programs on the market. Today I only use two: 1) Photoshop's PhotoMerge feature, which also allows an option to blend on the sharpest pixels from the overlapped areas of each panel. 2) Microsoft ICE, which is a freeware and is a snap to use has some fantastic save-as options like being able to save the panorama as Photoshop layers (PSD or PSB files).
Final Post Processing: Once I have the final stitch, I determine what to crop away and what areas to clone in if needed. I save the cropped image as a Photoshop file and begin making my global layers adjustments using a series of luminosity masks, HSL adjustment, etc. When I'm finished with the editing I save the final image as a 16-bit uncompressed TIFF and bring it back into Lightroom so that I can apply (non-destructive) output sharpening and vignetting (which is conditional).
Please use the comments form below for any questions related to this article.
It doesn't seem possible to re-focus for each panel but obviously your swamp scene proves me wrong!
Now think about that, I wonder if the HDR concept could be used as a depth of field tool? Like isolating the sharp focus pixels from multiple images of the same scene and layer them to a single image that has a sharp background, mid ground, and foreground?
In other words a change of focus, instead of f-stop, for each layer to be composited.
What a great post. Very detailed and helpful. I appreciate the complexities or planning for, capturing and processing such images on DSLR cameras. This is one reason I love my iPhone captured Autostitch option. Took most of the work away and keeps all the fun.
You have wonderful panos. Thanks again for sharing.
No comments posted.
Recent PostsLightroom Photo Merge DNG versus 32-bit Floating Point TIFF Reversing Order For Focus-Stacked Images Lens Selection, Focal Length, And Perceived Distance To An Object Maximizing Your HDR Photography Taking Control of Sharpening in Lightroom A Panoramic Photography Workflow A Method for Black & White Conversion