views:

70

answers:

3

So we have a photo like this

alt text

How to detect that a red wall has a white figure painted on it and that that white figure is a texture and than how to cut that wall from the picture? I need an algorithm for performing such operation programaticly (not by hand)

+1  A: 

Given that this is a difficult task for a human to perform, I expect it currently impossible to do so algorithmically. It is hard to compete with a visual processing system that has been over 200 million years in the specification refinement stage with coding errors punishable by death.

msw
agree. but if you still want to take a look at the state-of-art algorithms, try "computer vision" - example of research: http://www-cs.stanford.edu/groups/vision/documents/NieblesHanFerenczFei-Fei_ECCV2008.pdf
darlinton
@darlinton: I've seen that one=) But it's for videos\animations - not for static ones=)
Blender
i see... example for static pictures - http://vision.ai.uiuc.edu/?p=823
darlinton
how to solve it by your hands using photoshop http://www.dedestruct.com/2008/03/12/how-to-extract-people-and-objects-in-photoshop/
darlinton
A: 

There is no silver-bullet algorithm to do what you're asking in the general case. Computer vision is an one of the most active areas of research in modern computer science. The problem you pose is the holy grail of a lot of researchers.

The best you can get is manually messing with the image in Photoshop (or similar) using the magic wand tool. You currently need a human to decide whether you've selected the correct part of the image with the tool.

Alan
A: 

If you're looking for a simple algorithm that you can implement yourself, I'm not sure such a thing exists.

If you're doing research in the area, a good place to start would be "Modeling Textures with Total Variation Minimization and Oscillating Patterns in Image Processing" by Vese and Osher (see Vese's publications page). In particular, figures 14 and 15 show the results of applying their method to a similar image (although with higher-frequency detail).

celion