Long-Term Time Lapses in 3D

In my last post, I described how to create decent long-term time-lapse stills and animations from photographs taken with an ordinary handheld camera.  Now I’d like to ratchet things up a notch by extending the same techniques into 3D: taking pairs of photographs repeatedly from the same horizontally separated spots day after day, auto-aligning them to compensate for minor shifts in camera position, and averaging the results, either all together at once or in overlapping sequential groups.  You’ll need a pair of red-cyan glasses to appreciate the 3D effect in most of the examples I’ll be sharing, as was also true of my earlier post “Historical 3D Fun With Anaglyphs.”  You can easily order a cheap pair online if you don’t already have one.

Whether we’re attempting time lapse in 2D or 3D, we have different options to choose among depending on what aspect of a scene we want to emphasize.  Let’s first survey what those options are in 2D.  Maybe we want to present a frenetic whirlwind of sped-up activity, in which case we might take all the source photos, auto-align them, and show them in rapid succession, as with this sequence of daily pictures I took of our kitchen over a number of months:


Or maybe we want to draw out only the elements that remain constant from day to day, in which case we might create a single still image by averaging all the auto-aligned source photos together:

kitchen-time-lapse-average-stillOr maybe we want to achieve something in between by highlighting details that don’t necessarily remain static over the course of months, but that don’t jump about frenetically from day to day either—in other words, exposing gradual change over time.  Each frame in the animated GIF below is an average of multiple auto-aligned source photos in successive and overlapping groupings of thirty, such that time progresses at the same rate as it did in our first animation, but we’re always looking at the sum of about a month’s worth of images at a time:


The last approach is the one I focused on developing in my previous blog post.  But one thing I like about all three of the alternatives shown here is that it didn’t take any special gear for me to capture the images that fed into them—just an ordinary camera, with results processed afterwards on an ordinary computer.  It’s always a nice challenge to try to achieve cool effects without using specialized tools, and I try as much as I can to devise and describe techniques here at Griffonage-Dot-Com that the average reader could readily try out for himself or herself.

In the same spirit, I’ve also been experimenting with taking 3D photographs using an ordinary camera—snapping one picture and then moving my camera a little to the left or right before snapping another—rather than a specialized one with two lenses.  Here’s a selfie I took in a bathroom mirror to show how this works.  Notice the difference in the position of the camera in the left and right views.

3d0013 - P1340978-3d-smBut usually I haven’t been taking selfies.  Here’s a view of the fence along the western boundary of Bryan Park in Bloomington, Indiana:

3d0009 - P1340942-3d-smAnd a view through the top of a table at the same park:

3d0008 - P1340929-3d-smAnd an image from along a trail at Griffy Lake Nature Preserve:

3d0005 - P1230954-3d (detail)-smSo I was experimenting with these two things separately—time lapses and 3D anaglyphs—and thought I might as well combine them.

One of my time-lapse subjects has been a red buckeye tree on the Indiana University Bloomington campus which I’ve been photographing from both sides of a rectangular memorial plaque stuck into the ground in front of it, aligning my camera first to the right edge and then to the left edge.  Here’s a 3D anaglyph showing the plaque itself so that you can get a sense for the vantage points it offers.

red-buckeye-explanatoryThe “left” and “right” views in the next anaglyph are both averages of 173 images apiece, taken between August 2015 and August 2016 with my camera lined up with the sides of the memorial plaque.

campus-tree-yearlong-anaglyphAnd here’s a side-by-side stereoview scaled so that you can print it out and view it through a vintage stereoscope if you’ve got one:

campus-tree-yearlong-stereocardOf those two display options, the anaglyph is more useful when it comes to displaying 3D animations online.  But the process of producing it is more complicated for an animation than for a still.  The easiest way I know to create a single anaglyph in Photoshop is to overlay the left view on top of the right view, adjust it by going to Blending Options and unchecking the G and B channels under Advanced Blending, and then align the two images by trial and error until they yield a stereoscopic effect.  To handle the larger quantities of images required for animations, I’ve been manually aligning a single pair of the “left” and “right” images in an anaglyph, using those as templates for aligning the complete “left” and “right” image stacks, and then using a Photoshop action repeatedly to copy corresponding “left” and “right” images, paste them into a target document, adjust the blending options on the “left” image, and merge the results.

Here’s an animated GIF of just the left view, averaged in groups of ten source images.  (Download at full scale here [48.3 MB].)


And here’s an animated GIF of anaglyphs combining left and right views, again averaged in groups of ten source images.  (Download at full scale here [45.3 MB]).


That was my first attempt at a time-lapse animation in 3D.  For my second attempt, I chose a view of the street where I live: Tudor Lane in Bloomington, Indiana.  I’ve been capturing one sequence of photos with my camera lined up with the front of our mailbox (the left view), and one with it lined up with the back of our mailbox (the right view), generally taken once a day around 5:30 or 6 PM.  First, here’s a 3D anaglyph of the mailbox itself to give you a sense of the two vantage points.

mailbox-explanatoryNext, here’s a still based on 341 pairs of source photographs, or 682 photographs total, taken between July 23, 2015 and August 12, 2016:

tudor-lane-one-year-anaglyph1Next, here are all the individual pairs of source photos presented in reverse order (i.e., going backwards in time) as anaglyphs, aligned but not averaged (download at full resolution here [97.6 MB]):


And here’s the same sequence averaged going forward in groups of fourteen images (download at full resolution here [105.1 MB]):


I have plenty more source material on hand which I could prepare in this way, but the two sets of examples above should suffice for a proof of concept.  The techniques I described in my last blog post can indeed be extended into 3D to good effect.  Why not give it a try yourself?


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s