Fall 2018 CS194-26 Project4

Face morphing

Jieming Wei

Morph animation and mid face

Here, I morph my face to Obama's. I first labeled facial features using 46 points. Few more static points are also added on sides of the pictures. Triangulation are calulated based on the average facial points of the original and target photos. All pixals on the entire pictures should be inside of any of triangles since I added points on each sides of the pictures. For each transitional frame, for both target and original image, I calculate the linear transformation of each triangle from the picture to a weighted mid point and use the inverse mapping technique to get the interpolated value of each pixal. The bottom left one is an animation consisting of 45 frames of transitional picture. The bottom right one is the mid face, which is averaged picture of both origin and target picture morphing to mid shape.

Original face
Target face
Morph gif
Mid way face

Mean face and caricatures

Faces example morphing to average face (no smile)

I used FEI face database on this part. The dataset have both smile and non-smile face. First, an average shape of all faces data get calculated. Then all faces get morph to the average shape.

These pictures shows few example face that get mophed to the average shape.

example 1
example 2
example 3
example 4
example 5

The first image is the mean face of non-smile face. The second image is the mean of of all smiling face. The third image is mean non-smile face mophed to shape of my own face. To achieve this, I define my face feature following the database's way of facial factures and then used the moph technique described in part 1 to do the morphing.

mean face no smile
mean face smile
mean non-smile face mophing to my face

The first picture is my original face. The second shows my face morhed to average non-smile face. It makes more sense to morph to non-smile face instead of smile face because I am not a smiling boy. I did not see a dramatic change of my face after morphing probably bacause my face shape is close to average. The third picture is a caricature of me, which get computed by subtracting the difference of my shape and average shape from my shape.

My face original
MY face morphing to non-smile mean
My caricature face

Bells and Whistles, change ethnicity

Here, I change my ethinicity to white. I find a picture of average white male online and used it.

My face original
Average white male face

I tried morphing only appearance (left), morphing only shape (mid), and morphing both (right). Morphing appearance works the worst since it requires a perfect match between different feature positions. Morphing both gets me the best result. Eyebrow is not one of the facial features I defined. Therefore, they are not aligned.

Morph appearance
Morph shape
Morph both

Bells and Whistles, morph movie

I made a movie on morphing between different Real Madrid players. Hala Madrid.