When can a photo be trusted?

Fourandsix on TwitterFollow @Fourand6
« From Possible to Practical: The Gap between Research and Product | Main | Has The Alleged Iranian Spy Been Photoshopped? »
Tuesday
Oct082013

The Photo Of The Alleged Iranian Spy: A Followup

Last week I discussed questions surrounding the photo of an accused Iranian spy. At first glance the shadows in the photo seem somewhat odd, but a forensic shadow analysis revealed that the shadows are physically plausible. Here I show how this analysis can be further refined to provide even stronger evidence of this photo’s viability.

[Source: The Blaze]

In the above photo, the shadow of the telephone pole may seem at a peculiar angle as compared to the man’s shadow. Shown below is the result of a forensic shadow analysis used to determine if these shadows are physically plausible. The cast shadow from the pole, car, and man (red dots) and the attached shadows on the man’s shoulders (white dots), each constrain the projected location of the light source. Because these constraints have a common intersection (the black outlined region), the shadows in this photo are physically consistent.

We know from this analysis that the shadows in this scene are physically possible and that the projection of the light source is somewhere in the outlined black region (the intersection of the shadow constraints). Because this intersection is not a single point (and in fact extends upwards to infinity) it is possible that the shadows are in fact not consistent with a single light source, but arose from two different light sources each of which happen to be contained within the same region. As such, the smaller that we can make the intersection of the constraints the more likely it will be that we will detect inconsistent shadows.

As shown below, by simply adding two more constraints from the shadows on the man’s leg we can significantly reduce the constraint intersection size. This makes it that much less likely that this photo is a fake.

It can be challenging to narrowly constrain the projected light source, in part because the shadow to object correspondence can often be ambigous leading to broad individual constraints. As such, the more shadows that can be specified, the more narrow the viable region of the light source will be. This in turn will make it more likely to detect a forgery.

PrintView Printer Friendly Version

Reader Comments (3)

I think there is an other way to answer the question wether it is a fake or not.

Why would it be a fake, its just a picture, there is no special relation between the man and the place. So if there is no obvious reason for the picture to be a fake it probably is not a fake and even if it were, the man was photoshopped into another place, what would that matter?

The only unanswered question is of course: who is that man? But that is not something that can be solved by analyzing the picture.

October 8, 2013 | Unregistered CommenterEduard de Kam

Interesting point Eduard. I think your comment reflects a reasonable belief about the matter but hardly one that is conclusive or definite. Therefore it does not 'answer the question', per se, of authenticity versus alteration.

Indeed, I would suggest that the position you present relates mainly to the prior odds for the situation - that is, without considering Dr. Farid's forensic analyses, it seems extremely unlikely that the photo should be a fake. Prior odds, or pre-existing beliefs, are important when reaching a final decision about something like this. But, strictly speaking, prior odds are not relevant to the role or function of a forensic examiner. Their role is to evaluate evidence present in the material to determine its probability of occurrence under each of the different alternative propositions or hypotheses. Remember that, even though fakery might seem unlikely based on the idea that "there is no obvious reason for the picture to be a fake", that issue was still raised based on the idea that "at first glance the shadows in the photo seem somewhat odd".

As such, Dr. Farid's analysis and interpretation addresses information of a more objective nature derived from the image itself that can help in our evaluation of the situation. That information ultimately serves to change or shift our belief in the matter one way or the other. In this case, the shadow analysis provides strong support for the belief that the photo is not faked (and stronger than was provided by the earlier analysis) with little support for the alternative belief that the photo is a fake.

It is best to think of such information as a type of likelihood-ratio which serves to modify our already-existing odds through a multiplicative procedure. Granted, we cannot easily quantify either the prior odds (which do seem to be rather small to begin with) nor the likelihood-ratio (which strongly favours the photo being legit and not faked). But even lacking specific numeric values it is clear that Dr. Farid's analysis is very valuable in that the evidence shifts our belief strongly in favour of the one possibility and away from the alternative. In this case, the evidence derived from the analysis happened to support our prior beliefs but that doesn't happen all the time.

May 13, 2014 | Unregistered CommenterBrent Ostrum

That's an interesting technique. However, the analysis seems to miss one crucial factor: the apparent light source, which is the Sun, is at infinity (=very far). Are the shadows still explainable given that additional constraint?

This shadow analysis applies whether the light source is very far or nearby. More details about this analysis may be found here. -Hany

June 14, 2014 | Unregistered CommenterLadyslaw Wlodarski

PostPost a New Comment

Enter your information below to add a new comment.
Author Email (optional):
Author URL (optional):
Post:
 
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>