I am mainly interested in the understanding the techniques used for this and in algorithms for recovering the occluded objects from the acquired data. Can somebody recommend relevant references? Thank you in advance!
I think you should look in the direction of image segmentation and supervised classification.
There is a related but simpler question in RG (Atul Halmare Shri Guru Gobind Singhji Institute of Engineering and Techno...I'm working on overlapped character recognition but not getting success in segmentation of characters, can anyone offer any suggestions? ) that might be useful . It would appear that without prior additional data such as another view .etc .it might be necessary to make some assumptions for a given case to solve such problems e.g overlapping cells in microscopic images.etc.
Cheers
Please see (some wild suggestions below):
1 A very simple case would be "X" ie 2 line segments crossing. If included angle theta is>0 then there is only one common cross-over point. We can get this only if we make the assumption of continuity of the segments at the common point.!!
2.A possible (just a conjecture ) with objects in images is to reduce objects to closed line polygons and use solution of 1(e.g skeletonization). Another possibility is to use fractal geometry of objects in the image.
Cheers
Another suggestion: If you can assume that the occlusion has a spatial frequency that is very different from the image that has to recover, a filtering process in the Fourier space can be useful.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Algorithms