Monday, August 14, 2017

Detecting photo manipulation--the classic way

Before I write up the answer about the Yucatán.... 

I want to tell you about something fascinating I heard over the weekend. 

RadioLab is one of the best podcasts going.  It covers a broad range of topics in science, education, environment, health, and life-in-general.  Their podcasts are incredibly well written and produced.  To my mind, they set the standard for what a great podcast can be.  

Usually I listen to their casts as I'm running errands on the weekend. On Sunday, I listened to a RadioLab podcast about a search process that's of real interest to SRS readers.  

Truth and Cannonballs (22 mins) is about the quest of documentary filmmaker Errol Morris to understand how a famous pair of images from the Crimean War (by 19th century photographer Roger Fenton) were manipulated.

Here's the pair of photos side-by-side:  

It's pretty clear that at least one of these pictures was manipulated by moving the cannonballs around.  The question that consumed Morris was "which of these was first?"  

That is, did Fenton first photograph the road WITH the cannonballs, and then move them away--OR--was the road empty, and he placed the cannonballs there for photographic effect?  

The podcast has a great discussion about what motivates someone to pursue a SearchResearch question like this, and then what he did in order to figure it out.  

tl;dr -- a colleague used a version of the blink comparator (that we discussed a few weeks ago) to find some stones that moved between the two different versions.  That was enough to say with high confidence that clear-road version was first.  (The insight was to realize that the stones all moved downhill, suggesting that they were accidentally kicked as the photographer moved the cannonballs into place.  After all, it's very unlikely that all of the stones would move uphill!)  

Listen, and enjoy.