In previous article, we mentioned the pioneer of the face recognition; Bledsoe and his researches.
In this article we’ll talk about how this researches progressed to the phase of automatic face recognition. In other words; “70s,80s and early 90s of face recognition”.
70s; Getting Practial
In 70s, science people A.J. Goldstein, Leon D. Harmon and A.B. Lesk used a more practical approach to Bledsoe’s RAND Tablet using technology and managed to get better results.
They listed 21 criterias such as hair color, lip boldness to automatise matching process. But measurement and locations must be calculated manually as in the Bledsoe’s system. Unfortunately it was taking huge time.
80s and 90s; Math and Computer Science Joining Forces; Eigenfaces Algorithm
In 1988 mathematicians Michael Kriby and Lawrence Sirowich used linear algebra princible and developed a system based on it. They named this approach as Eigenface.
This principle was simplifying face photographs and summarizing them with parameters with less data. Count of these parameters were less then one hundred.
Eigenface algorithm was calculating the average of these values, then comparing the difference between the average for each parameter.
In 1991, Matthew Turk and Alex Petland progressed the Eigenfaces algorithm. They also found the way to find the faces in a bigger picture. This was the milestone for the way of automatic face recognition.