Project 3: Face Morphing
and Modeling
by Patrick Callahan
I used IDL to define the 43 necessary points on the faces. The .txt file of the pixel locations is available here. IÕm not sure what others were using for their x, y positions, I just used the actual pixel indices.
First, the morphs:
|
Start |
Middle |
End |
|
|
|
|
Obviously, the beard makes for a bad morph to someone without a beard, so I thought IÕd see what it looked like with three pictures of me with different beard sizes. I wanted to see if I could ÔgrowÕ a beard with face morphing. I think it came out pretty well; the only problem is that the second photo was from belowÉ
The Mean Face:
I calculated the mean face for all the students in the class and Prof. Efros.
Then I calculated the mean male and female faces using the mean points of each.
|
|
|
|
mean female face –
female points |
mean male face –
female points |
|
|
|
|
mean female face –
male points |
mean male face –
male points |
Bells and Whistles
Caricatures, feminizing, masculitizingÉ
I produced caricatures using the average class, female, and male
points. IÕll just show the first couple as there are too many to show for the
entire class.
|
|
Original
|
|
|
|
Mean
|
|
|
|
Feminizing
|
|
|
|
Masculitizing
|
|
Principal Component Analysis
I performed a PCA of the faces all transformed to the mean shape. I used a built in IDL singular value decomposition function. Below are the first 9 of the 32 Eigen vectors. The first is the mean face, the second appears to represent hair, the third the lighting above the eyes and clothing, the fourth foreheads, the fifth teeth and glasses, and so on. ItÕs pretty difficult to tell after the first couple.
|
|
|
|
|
|
|
|
|
|
|
|
I also did a morph between my brother, me, and my mother.
