Simple Face Detection
So far so good. So, Skin Recognition is done .. however there might be objects in the image which have the same skin color but arent faces.. what bout those .. so our detection isnt complete, we need to detect faces in those skin areas of the new image I.
To cope up with this we need to the image I on various scales at various points in order to locate a face in the whole image. Scanning is done in all directions - left to right, top to bottom, starting a width n height of the image n reducing it to the minimum size unit. Now this can be a lengthy process n is very processor hungry… it is suggested to scan only at center of all locations in the image I which has a non zero Sij.
Now how do we determine whether an area is a face or not. This is done by subtracting a face template F from the region of the raw image A. The face template is calculated by averaging the pixel values over a number of face example images. After doing this in all directions the resulting image is called an Eigen Face.
Eij = |Aij - Fij|
This Eigen Face is then fed to the input units of an Error Back propagation Neural Network which is already trained with a database for determining a face from a non-face.Yea things have suddenly started getting over us .. isnt it .. Well if you are still with me then onto to the next part of it … Yes the last n final and the most important part of it … Recognising Faces …
Determining whether it is you, me or someone else .. Riya's Face recognition does this on the fly when u manually tag some of the faces n click on to Automatic Recognition Feature … So lets get started as to how is that done n get our hands dirty, punching it out.
Leave a Reply