Matlab Image Processing Help (Image Segmentation)

Thread Starter

peep2k

Joined Nov 5, 2009
2
Hi everyone,
I would like to use Matlab to perform human recognition by detecting human head in images. For e.g. on this image:

http://img46.imagevenue.com/img.php?image=90238_1_122_126lo.jpg

As you can see the most visible portion of human from this view is the head. The problem im having is that how can i completely extract human head from the image like I have use canny filter to detect edges of the image and it is able to detect the boundary of human head but along with extra pixels which is not under my consideration like shoulders, hand and some extra pixels due to background noise.
Eg:
http://img229.imagevenue.com/img.php?image=49208_1canny_122_92lo.jpg

The result i wanted is something like this(i have use adobe photoshop):
http://img161.imagevenue.com/img.php?image=90243_1cannyerosiondilation_122_595lo.jpg
with only the shape of human head is visible

I would appreciate some help into how to transform (script, functions) images in Matlab to do so.


Thanks everyone.
Cheers,
 

steinar96

Joined Apr 18, 2009
239
I'm not too familiar with facial recognition but usually it's done by rather complex systems i think. A friend of mine for example is trying to train support vector machine (supervised machine learning) to do facial recognition.
 

Thread Starter

peep2k

Joined Nov 5, 2009
2
Yesterday 12:19 PM steinar96 I'm not too familiar with facial recognition but usually it's done by rather complex systems i think. A friend of mine for example is trying to train support vector machine (supervised machine learning) to do facial recognition.
Sorry but Im not asking for facial recognition , more specifically , all i wanted is to detect number of circular objects in the image even if it detects some of them which are not human head but the point is that it should automatically removes all the background noise and objects which are not circular in shape leaving only the circular objects in the image
Thanks.
Cheers
 
Top