eigenface

use first 5 eigen vectors.

on top, add eigenvector to average. It's like S(+ + + + +).

on bottom, subtract eigenvector to average. It's like S(- - - - -).

on top, add x value, subtract y value, and add z value to avarage face. ( +x, -y, +z )

it looks like weight loss.

on bottom, this is average face.

note: first eigenvector is multiplied by 1000.

sample face image

highest eigen vector times 50 added to average face (right) compare to average face. (left)

so, it's like avg[i][j] + (50 * eigen[i][h])

Because values of emelemts in eigen vector are so small, it will not show huge change unless I multiply by large number such as 1000.