How to differentiate a real person and a picture of him in an image not in video stream?

Is it possible to differentiate a real person and a picture of him in an image and can it be recognize that in image which is real person and which is a picture? And if it is possible then what algorithm can we use for this purpose?

Answer 1

[Image Analyst]

Possibly. Are you using a thermal camera? If so, look for heat. Otherwise register the images from several seconds apart and subtract them. If it's a real person, there will be some differences. If it's a picture, there will be little difference - basically just a small amount due to video noise.

[Shakir Ishaq]

Thank you so much for helping. If you don't mind I need a little more detail about this. For example I have an image in which someone wear shirt on which there is a picture of someone or there is a picture of someone on the wall which is also the detected by the system, then how these will be differentiate. Here the picture is:

How we will solve this problem...?

[Image Analyst]

For that particular image, I think you'd have to do a color segmentation. The face on the shirt is noticeably redder and could be detected and filtered out. But what you do for that image may not work for the next such image. Luckily I don't think you'd encounter such an image as this often. Even if you did, what's wrong with identifying the face on the shirts?

[Shakir Ishaq]

I have problem not only from this image, I want to create a such application which told us that which one is a real and which one is a picture in the image. Like given in the picture.

I mean that if someone show a picture of someone to to bypass or for other purpose and the system not accept such things.

[Walter Roberson]

The systems for detecting pictures being presented instead of people, generally use one of a small number of methods:

1. additional camera(s) to get a 3D view, since flat pictures have no depth; Or
2. zoom in really far to look for pores and other features that pictures do not have; Or
3. careful measure of location of features over a period of time, reflecting the fact that people do not stand perfectly still and that facial features change slightly as people breathe; Or
4. use of additional sensor modes such as infrared (pictures have wrong heat distribution) or ultrasound (distance measures) or chemical sensors (such as CO2 or ketones)

What the systems do not do is try to make a determination based upon a single image, the way you are trying to do.