tomy23
November 5th, 2010, 02:14 AM
I have a face recognition project ,
I finished the step of convert RGB image into binarization image (by thresholding)
now I have a black-white image, and I need to segment my image in to objects so i will be able to exract features
Is my steps true??
Can anyone help me to give me algorithms for segmentation ( and src code if there)??
BioPhysEngr
February 5th, 2011, 01:53 PM
Hi,
One of the more classic libraries for doing this is OpenCV. It is available at http://opencv.willowgarage.com/wiki/ which mostly works with C++. There appears to be a .NET wrapper at http://code.google.com/p/opencvdotnet/. I've used OpenCV before and it has worked well for me.
These rely on (if I remember correctly) Viola-Jones object detection which you can read about on Wikipedia: http://en.wikipedia.org/wiki/Viola-Jones_object_detection_framework. The classic paper on that topic is available at: http://research.microsoft.com/en-us/um/people/viola/Pubs/Detect/violaJones_IJCV.pdf. Note that they use face detection as part of the example.
You should probably be aware that face detection (i.e. 'is this a face?') is a lot easier than face recognition (i.e. 'whose face is this?'). There is a general treatment on Wikipedia: http://en.wikipedia.org/wiki/Facial_recognition_system and some otherwise good monographs on the subject. One that I looked through when I was an undergraduate (but did not do anything serious with) was: Face Processing: Advanced Modeling and Methods http://www.amazon.com/Face-Processing-Advanced-Modeling-Methods/dp/0120884526/ref=sr_1_1?s=books&ie=UTF8&qid=1296935421&sr=1-1. I wouldn't necessarily recommend you buy it, but if you have an academic library (university, community college, etc) nearby, they might have it available for reference or borrowing.
Long story short, it's a complicated but extremely interesting problem.
Hope that helps point you in the right direction!