Facial recognitionn Software - How does it overcome different angles & distance?

Factual Questions
1

I’m not rich enough to buy facial recognition software. It’s not priced for the home consumer at all.

I have been trying to implement it in Photoshop. I’ve seen it used in various History Channel documentaries to test whether older people claiming to be Jesse James or Butch Cassidy were lying. Basically, They try to overlay the two faces and confirm the eye sockets , Nose, and mouth match. They can do it with two photos or a skull and a photo. Those adult features won’t shift with aging. With Photoshop that’s pretty easy with layers and adjusting the transparency. I rescale the photos until I get as close a match as possible.

However, I’ve quickly learned there’s more to this hat trick. (1) It’s very unlikely that you’ll get two pictures where the people hold their heads in exactly the same angle. (2) Or one shot will have the head tilted more than in the other photo. (3) Scaling still doesn’t compensate entirely for the difference in the distance a photo was taken from. A close up head shot just won’t scale out the same as a full body shot from seven feet away.

Since I never have succeeded in getting the two photos perfectly aligned, then I really can’t compare eyes, nose and mouth to see if they align.

Oh wise Dopers. Can someone kindly point me towards a manual method of using Photoshop to compare photos? The commercial software claims it can pick out a face in a Superbowl crowd and compare it to known felons with warrants. I’m not that ambitious, but would like to play and compare photos of people when they were young and old.

2

The idea, at least, is that it creates a three-dimensional grid. Sort of like 3-D modeling for movies like Toy Story…just done in reverse. Once you have the grid, you can rotate it, tilt it, expand it, etc.

It’s one of those technologies that’s getting better all the time, even if it may not quite yet be “good enough.” Like voice recognition. Ten years ago, V.R. was awful. Today, it can work fairly well – well enough to do some basic phone-tree sorting. “Press or say ‘five’ if you want to cuss out a live affiliate.”

3

Google picassa does a fairly decent job of face recognition. It seems to be doing a pretty good job at automatically tagging the thousands of pictures I have taken over the years. Not 100% but 80% to 90% if the photo has people facing the camera. The software is free and is useful for organizing photos.