Tag Archives: computation

More Computational Photography

As an Electrical Engineer and a long time photographer, I’ve been interested in computational photography for a while.  You can read some of my  earlier posts on the subject at this link. Here’s Wikipedia’s definition:

…”digital image capture and processing techniques that use digital computation instead of optical processes.”

I’d change this slightly from”instead of” to “instead of or in addition to”, but that’s a small quibble.

Moore’s law keeps driving the computing capabilities in phones and cameras ever higher and it’s fascinating to see what companies do with the extra potential.

Towalga River below High FallsTowalga River below High Falls, Georgia. Composite – iPhone default live view and Long Exposure, merged in Photoshop

I made the image above on our recent trip using an iPhone 8+ and Apple’s “Live Photo” mode. After I made the photo, I edited it on the phone and enabled the “Long Exposure” effect. The Long Exposure effect of blurring the moving water is computed from ~3 seconds worth of frames that Live Photo captures.  This is Apple’s web page explaining the feature.  Here’s what it looked like before I changed the mode – it’s not nearly as photogenic:

Towalga River below High FallsDefault iPhone live view

Until iOS 11 added this feature, I wasn’t too interested in Live Photos.  Now, I’m watching for places to use it. You can get a better image with your high-end camera and traditional optical techniques, but this is easier and a lot of fun to play with.

Here are some hints:

  • Pick a suitable subject:  moving water, traffic on a road, blurring people in a crowd, etc.
  • Motion blur with a traditional optical approach requires a slow shutter speed – either low light or using filters.  Since computational methods works by processing multiple frames, you can use it in bright light without filters.
  • Apple says it works on their newer phones (6+ and later).  You’ll need to have iOS 11 (or later) installed.
  • The Long Exposure effect has to align Individual frames and then crop where there’s no overlap so you’ll lose pixels around the edges.  Ideally, use a tripod – but that sort of defeats the idea of pulling your phone out of your pocket, doesn’t it?  Just hold the phone as steady as you can to minimize cropping.
  • Make several exposures and pick the best one later.
  • Long Exposure resolution seems to be lower than default iPhone photos.  This isn’t a huge problem for the moving parts of the frame – they’re supposed to be blurry.  For the static portions, you can load both versions into layers in Photoshop and use masking to paint in higher resolution where you want it.  I did this for the first photo above.
  • You can set a Long Exposure photo as your wallpaper.  You’ll see the static Long Exposure version until you press on it from the lock screen.  Then it changes to show the three-second animation – cool!

I hope Apple enhances this in future updates.  It’d be good to have some control over the blur effect.  3 seconds is nice, but some subjects will look better with less (or more?).

iOS 11 includes other updated computation photo capabilities (e.g. portrait lighting) – but that’s a subject for another day.

Photography’s changing fast – it’s a wonderful time to be a photographer, isn’t it?  In today’s digital world, many advances are likely to be computational and not optical.  Keep up – don’t be left behind!

Thanks for stopping by and reading my blog. Now – go compute some photos!

©2016, Ed Rosack. All rights reserved.

Pocket Computational Photography

If you’ve followed this blog for a while, you may have seen my earlier posts on computational photography.  If not, you can review them at this link:  https://edrosack.com/?s=computational+photography.  The term refers to using software algorithms to supplement or replace optical capture processes.  Common examples are multi-frame panoramas, focus stacking, HDR processing, post capture focus, and other techniques.  You can read more about it at this link on Wikipedia:  https://en.wikipedia.org/wiki/Computational_photography

As phone capabilities increase, their computational photography power is growing.  Camera phones have long been able to do on the fly panorama and HDR capture.  And here’s an example of a new capability that arrived on the iPhone 7+.

BokehBokeh

Apple calls this “Portrait Mode”.  It’s available in Beta on the iPhone 7+ in the latest version of IOS.  Since the 7+ has two cameras separated by a small distance, it provides the info necessary to compute a “depth map” of pixels in the frame.  The software uses this to selectively blur pixels based on distance to add a “Bokeh” (shallow depth of field) effect that helps with subject isolation.  For comparison, here is the non-computed version of the image.  You can see that the background looks very different.

Original
Original

All isn’t perfect.  The algorithm has problems around small features at the boundaries.   Look closely at the next frame and you can see blurring issues at the edges of the reed.

Phone output
Phone output

The processing blurred parts of the reed that we wanted sharp.  For the first photo above – I cheated and used Photoshop to correct the problems.  Maybe in future versions the software will be better.

Here’s one more example.  This is Lynn, rocking an election day t-shirt.  First, the portrait mode version.

Lynn - original
Lynn – portrait mode

And finally, the original.  In this case, the software did much better, with no obvious blurring issues.  These two are straight out of the camera with no processing on my part.

Lynn - portrait mode
Lynn – original

It’s fascinating how photography and computers are merging.  For someone who started out programming a large room sized Univac in FORTRAN with punch cards, the power and ability that fits in my pocket is just stunning.  I’m glad to have it with me.

What can they possibly think of next?  Do you use computational photography techniques?  Do you like or hate them?

Thanks for stopping by and reading my blog. Now – go compute some images!

©2016, Ed Rosack. All rights reserved