Differentiating real time object and a object in a image.

How to differentiate between a object, say a ball, shown in front of a camera and a photograph of a ball shown in front of the camera using matlab?

Answer 1

[Image Analyst]

I don't think you can, unless there's more information, such as the ball is surrounded by a rectangle that indicates the edge of the photo print, or you have stereo views that could be analyzed to find flat objects, or you can resolve dithering, or you have multiple real balls and a photo where the photo doesn't have the same appearance as the real balls, etc.

Answer 2

[Walter Roberson]

In some cases, you could handle it through an analysis of lighting.

In the case of the photograph, if the lighting of the flat photograph is point-wise (e.g., incandescent) or linear (e.g., florescent cylinder) then the lighting will effect a change in the intensity of the photograph, with added brightness closest to the light source and fade further away from the light source, and this intensity change will not vary with the content of the scene.

In the case of an actual object, the lighting would change across the object, possibly even with shadows being produced.

One cannot, in theory, differentiate between an actual scene and a picture of a scene that is isolated from external light sources and the picture projected towards the camera (a technique often used for animation.) (The "some cases" I referred to above include only cases where the photograph is not isolate from external light sources.)

Unfortunately I do not know how to perform lighting analysis.