Recognizing Image Manipulation

breezebongAI and Robotics

Nov 6, 2013 (4 years and 8 months ago)


Recognizing Image Manipulation

Fake "photographs" can now be produced by using "MS Paint", “Adobe PhotoShop”, “Corel
Draw” and many other image
processing software to rearrange, re
color and transform a scene.
The same software can combine fragments of di
fferent images into one new image. Other
software can generate synthetic photo realistic pictures by applying sophisticated perspective
projections and shading to digital models of three
dimensional scenes
a technique commonly
employed to present architect
ural projects and create Hollywood special effects.

Questions to ask when looking for inconsistencies

Do all the objects in the scene seem to be in the correct perspective?

Does the perspective foreshortening of a surface seem consistent with the spatial

orientation suggested by its shading?

Do indicators of time, such as clocks and shadows, all show the same moment of

Do some objects seem surprisingly light or dark in relation to their surroundings?

Are inserted objects betrayed by lack of s
hadows or by shadows cast at angles different
from those cast by other objects?

Are there shadows that do not seem to be cast by any object?

Are shadows and specular highlights consistent with the assumptions about locations of
light sources?

Do unexpec
ted discontinuities in the background of the scene suggest that objects have
been deleted from the foreground?

Do shiny surfaces display the expected reflections of other parts of the scene?

Are surface intensities appropriately modified by diffuse inter
reflection effects?

Are there plausible texture gradients as surfaces recede?

Do geometric perspective and atmospheric perspective convey the same depth

Is there a consistent gradient of sharpness from some focus plane?