This project morphs one image to another. We first define the correspondent points on both images. Then we find the mid-way face, and warp both images towards the middle. We also cross-dissolve the images. By repeating the process, we produce a morph sequence.
I attempted to morph my friend with her favorite celebrity.
original mica
original celebrity
warped mica
warped celebrity
mid-way face
morph sequence
Photo Credit: Mica and her favorite celebrity.
I used all the 400 sample images from http://fei.edu.br/~cet/facedatabase.html to get a mean face.
average
sample 1
morphed 1
sample 2
morphed 2
sample 3
morphed 3
sample 4
morphed 4
me
average
me morphed into average
average morphed into me
I produced caricature by scaling the difference between my face and average face by various factors.
difference * 0.5 + average
difference * 1.5 + average
I attempted to morph my friend with her guide dog. Some flaws in the sequence occurred because I didn't label enough correspondent points / some of the points don't match nicely.
original tiff
original helene
warped tiff
warped helene
mid-way face
morph sequence
Photo Credit: Tiff and Helene
I also tried to make Mica more Chinese.
original mica
average chinese
100% shape warp, 100% color change
100% shape warp, 50% color change
100% shape warp, 0% color change
50% shape warp, 100% color change
50% shape warp, 50% color change
50% shape warp, 0% color change
0% shape warp, 100% color change
0% shape warp, 50% color change
0% shape warp, 0% color change
Photo Credit: Mica and average Chinese