Facial Recognition Technics

Facial Recognition systemTraditional
A few facial recognition algorithms differentiate faces by extracting features or landmarks, from an image of a person's face. For instance, an algorithm might analyze the relative positioning, size, and shape of the nose, cheekbones, eyes and jaw. Those features are used then to search for some images with similar features. Other programs normalize a gallery of images with faces and then compress this face data by saving the data in these images which is useful in face detection. A test image is compared then with the face information. One of the first successful systems is rooted in template matching techniques used in a set of typical facial features, giving a kind of compressed face presentation. Recognition algorithms could be divided in two main approaches - the geometric, which goes for distinguishing features, and photometric - a statistical approach which distills the image into values and compares the values with a database of templates to eliminate any variances.

3 Dimensional recognition
A newly developed trend, claimed to provide improved accuracy, is the three-dimensional face recognition. The technique uses 3D sensors for capturing information about the size and shape of the face. The information is used while identifying distinctive features on the specific face, like the contour of the nose, eye sockets and chin.
One advantage of the 3D facial recognition is it is not influenced by changes in the lighting like with other techniques. It could also identify faces from a number of viewing angles, even a profile one. Three-dimensional data points from the face vastly improve the accuracy of facial recognition. Research in 3D is enhanced by the creation of sophisticated sensors which do a far better job at capturing 3D face date. The sensors function by projecting structured light on the face. Up to twelve or more of the image sensors could be placed on a single CMOS chip – every sensor captures another color of the spectrum.

Skin texture analysis

Another emerging trend which uses the visual particularities of the skin when captured in standard scanned or digital images. The technique, named skin texture analysis, transforms the unique patterns, lines, and spots visible in a person’s skin into mathematical space. Tests have proved that using the additional skin texture analysis increases the performance of recognizing faces by 20 or 25 percent.


The measurement process

The first part in the facial process is to acquire a real image or two-dimensional one of the person. The system determines then its alignment based on how the nose, and the mouth are positioned. After alignment, it makes a template and goes on to match it against the other images in its database.

In general the system measures the face in the following order:
• The width of the Nose
• The depth of the Eye sockets
• The distance between the eyes
• The shape of the Cheekbones
• The length of the Jaw line
• Anything else worth taking note of