Given that some software nowadays is so powerful you can easily stitch images together for a really wide view (either with panorama mode on camera or something like photoshop), is there still a point in bring ultra wide angle lenses on a trip?
Great question. A little over a year ago, I bought an ultra-wide (10-24mm f/3.5) lens with an eye toward landscape shots and quickly saw that generally, I can stitch images taken on a longer lens and produce more satisfactory panoramas. So, as you ask, what's the point of an ultra-wide?
Well, to answer it, the best approach is probably to discuss what an ultra-wide does, so you can decide how to make it work for you. I'll include a couple examples - not the best examples of my best work, but that will make my point.
First, an ultra-wide gets LOTS of stuff onto the same size sensor. This is why it fails for panoramas, and was never an excellent choice for them. It doesn't actually produce a bigger picture - it just makes stuff smaller.
Second, it tends to distort so that things on the edges look much bigger than they really are, proportionately to how the other things look. The wider the angle, the more this is true. Things in the middle tend to look normal, things on the edges, not so much.
Thirdly, it tends to exaggerate proportions. Whatever I want in the foreground, I have to get real up close to - like, within INCHES - everything else looks miles away. This is why it can fit big things in up close.
If you just want a panorama, then don't bother with the wide-angle. In my experience, the distortion means cropping and photoshopping and you end up with an image that captures just a bit more than a longer lens (say 24mm or 35mm), so what was the point?
But if you want to be able to get cool architectural features or interesting shots, then on vacation or anywhere, bring your ultra-wide.
I squeezed in a frigate from about 5 meters:
In this example, I am able to squeeze the whole church in (the tower is about 200 ft tall) by zooming way out to 10 mm and sitting inches from the family. Dad didn't think I could possibly get all of them in, let alone the church.
Because of the distortion, it isn't great for standard portraiture, but can be great for fun shots, like this one:
And cool special effects with the sky and clouds:
You can do things with a wide angle that can't be done with photo stitching. This photograph could not have been stitched; whilst I had time to take a few shots, I would never have had the chance to stitch it together.
Also the wide angle has a distortion effect, and this can be used for its specific composition effect and to draw attention to something in the foreground. The much greater depth of field it offers gives many advantages; candid photography and landscape scenes, for example.
If there is anything moving in the shot, then there isn't a substitute for a good wide angle lens. In addition, the difference in angle of the lens is going to result in a characteristically different feel from multiple shots at a longer focal length than a wide angle lens would have given on it's own unless you use a specially built mount that can rotate the camera body while keeping the center of the front element in the same location and then adjust for distortion caused by the lens.
There is also the simple time consideration that taking one shot with a wide angle doesn't require stitching and a wide angle lens really doesn't take up that much room, even for something like an L series lens.
For distant scenery, you are right you can stitch.
the time to process multiple images. Instead of 50 wide angle shots, you have 150 to stitch together. You'll end up with larger files, so more pixels to work with, but bigger files.
Ghosting - clouds, tree limbs, or people are moving may make seams problematic
Foreground interest. Most good WA shots have nice foregrounds - these may be difficult to capture and stitch without looking distorted.
Stitching is not mathematically correct. It is a 2D technique that works with image data. It simply warps the pictures to make them overlap without regard for perspective. Therefore it is confounded by perspective shifts when the camera view is rotated. For instance, lines which appear parallel when facing in one direction might converge in when the view rotates. There are wide-angle lenses which preserve straight lines, everywhere in the picture. When the picture is projected onto a flat surface (usually the case) and viewed from the correct distance relative to its scale, the viewer sees every object at the same angle as in the actual scene.
Objects and living things move between two shots. A wide angle lens captures one moment in time. You cannot stand at the edge of a soccer field during a game, take several shots, and then stitch them together to make a single wide-angle action shot.
Stitching is post-processing. Post-processing downgrades the data. The better you capture the image in the first place, the less post-processing you have to do. Ideally none.
Post-processing requires time. No optimization of the stitching process can match the efficiency of pointing and shooting once. If you want to produce a lot of wide angle pictures, it will be far less effort to just shoot them.
About point 1: it is possible to do a mathematically correct job. You need a linear lens, and the stitcher has to know the angles at which the pictures were taken so it can do the backward projection and then transform it into the perspective of the virtual viewer, and then resample to the combined wide angle picture. Take Google Streetview for instance. They take pictures with lenses that are mounted at 90 degrees, and the must also be a particular type of lens. The pictures then just have to be projected on the inside of a virtual cube, and then from there the cube is rendered, thereby resampling the pictures to re-create the view from the original perspective. Stitching would be used only to smooth out any minor artifacts in the overlap, not to do any major warping.
About point 2: you could shoot multiple view simultaneously with multiple cameras. Again, Google Streetview does exactly that. So at any one point in the data, you can look around 360 degrees (as well as up and down) and it's all consistent.