Time-lapse Geekery: Manual Panning and openCV
Emma and I went back to the Jungfrau region last week-end, and I experimented further with time-lapse photography. I discovered that I could make really cool panning shots by using a combination of the continuous exposure mode of my camera (which takes shots at regular intervals of maybe half a second) and video stabilization technology. The main challenge to make the shots look good is endurance: you have to hold the camera up, keep your finger firmly pressed on the trigger, and move very slowly and as continuously as possible. I accumulated on average 200 shots every time I did this. Then, you render the time-lapse into a video, and run it through a stabilizer. I’ve been using the stabilizer in iMovie, and it does a fantastic job. Who needs a 1000$ motion control system?
I found out I could do that largely because of my forgetfulness. Emma and I went on a hike to Bachalpsee, and I really wanted to take a time-lapse of the clouds moving over the Eiger being reflected in the lake. I had no luck: I forgot my intervalometer (and I don’t want to install Magic Lantern while my guarantee is still up), and there was enough wind to ruin the reflective surface of the lake. That’s without mentioning the obnoxious tourists who were just walking in front of the camera.
So I tried something else instead. I tried snapping pictures manually, at more or less regular intervals, while panning the scenery. Both sequences featuring the Bachalpsee lakes you see in my new time-lapse below were shot like this. I had no idea what would come out of it, and I was really impressed by the final result. It’s only later I realized I could use continuous shooting to achieve the same thing.
We were based in Grindelwald, where there is little light pollution. In the video below, you will see a nice sequence of the Milky Way coming into view which I got on the night of our arrival. The second night we stayed, I asked the hotel manager if I could leave my camera overnight on the balcony behind the hotel in order to get a better view of the mountains with the Milky Way flying over them. Then I mistakenly left my camera in automatic mode instead of manual… I felt bad all day about it. I got no stars. Still, I got some cirrus-type clouds over the Eiger, so not a complete loss, but the flickering caused by the changing ISO during sunrise is really annoying, rendering most of the footage unusable. Magic Lantern can correct for that really well apparently.
On a slightly different topic, I discovered this piece of open-source software called openCV, which stands for open Computer Vision. It’s a battery of tools (which comes with a fantastic python API, yay!) that can be used to do really advanced image processing like edge-detection and facial recognition. It boils down the treatment of images to operations on numpy arrays, and it makes turning a sequences of images into a video really easy. I now use openCV as the main engine in my scripts to do time-lapse preprocessing and composition. It’s also 10 times faster than using PIL. Look out the scripts here, it’s surprisingly simple.
Here’s the final result. Enjoy!
