Jason MacAllister’s Morph Extravaganza
For Computational Photography, 15-463 (Project 2)
Photos and point sets that were used for my experiments can be found aqui.
The 61 frames generated for class morph can be found here.
The first part of this
assignment was to generate 61 frames of a morph between myself and another
student. This turned out a little lika thisa:
Sorry about the borders,
these were generated in the “learning phase” (I still had no idea what I was
doing). These frames were all
approximately 300x300 pixels, allowing a generation time of about 30 seconds
per frame. The rest of my images are
usually 768x1024, and they took about 10-15 minutes
each (ouch!).
Mean (as in average) Face
Here’s the average face of my
class:
Our teacher referred to this
as the “average CMU undergrad.”
Here’s the average face
generated from 18 total donors (basically my class plus a large amount of my
family):
Here’s just the males (10
faces): And
the females (7 faces):
This is my face morphed to
the average geometry: And this
is the average face in the shape of mine:
Family Study
Next I looked at similarities
within families.
Here’s my family:
Aren’t they the sweetest?
Now, let’s play. Here’s the average:
Nothing like face averaging
to bring out the similarities in a family.
Here’s the men (me and my
dad):
Apparently I’m going to look
exactly like my dad when I get old- I mean older… It appears that I have no distinguishing
features, and the average just looks like a younger version of
Here’s the women:
And here’s the average of my
mom and dad. By far the most disturbing,
it really doesn’t indicate what their children look like as much as I had
expected. Mostly just looks like a
bearded mommy. It does bring out the
similarities in my parents’ faces, however, as several people have pointed out.
Next I averaged the faces of
my wife and myself, in the hopes of determining what our children will look
like.
Remember that picture of
Amanda? Well, here’s our offspring:
Maybe we should rethink that
whole children thing. Looks like a
possessed adolescent boy wearing suspenders.
Maybe a combination of the
wife and, say, the family dog would be more appealing:
Or perhaps me and my iguana:
Oh heck. Let’s try the two pets together:
THAT is crazily awesome. I had to remove the background (as in with
the Amanda-dog) cuz it was too confusing.
Enough torturing the MacAllisters. Let’s
see how the in-laws fare!
Here’s the rest of Amanda’s
family:
(this is the best picture I could
find of Zack, just ignore the other two, so I’m only pulling his face from it)
So here’s the family average:
They don’t seem to mesh as
well, especially since some have their mouths open and some don’t. Also, we get that weird collage effect around
the borders because of the differences in picture context. But you gotta love that bow tie!
Let’s look at just the men:
Apart from the strange
spirits floating around the head, this turned out to be a pretty good
combination. Those dang lips though.
Here’s the women:
Finally, here’s the average
male and female face from both families combined:
And the average of those two:
Caricatures
Here’s me with the deviation of my face from the average
amplified (distinguishing features become more pronounced):
…and I hate it. If you think about it, this is kinda the opposite of what I just did above (morphing my
face to the average geometry). Now I’m
morphing my face away from the
average geometry. So it makes sense that
my face is now skinnier, as opposed to fatter in the previous image. This image is exaggerated by one full
magnitude of deviation.
Here’s the same caricature
concept, only implemented with skin texture/color instead of geometry.
At first glance it seems like
it’s only higher contrast or brighter, but it’s actually only the parts that
are different from the average face
that are brighter. The redness of my
lips seem to be emphasized a lot, as well as my green
t-shirt, and any blemishes in the skin (because Mr. Average is OH SO PURE.)
Here’s my wife, Amanda:
And here’s her caricature
(geometry and texture):
As disturbing as it is, I’m kinda impressed that it works. She’s smiling a little more than the average,
so now she has a huge grin. And it
really emphasized the narrowing of her head width from top to bottom.
Also, here’s the texture
deviation from that caricature. This is
what’s added on top of the regular image to emphasize differences.
You can see the shine spots
that were brought out, as well as the brightness in her eyes.
Here’s some other caricatures of people you’ll recognize:
Appendix A
Other combinations:
+
=
Appendix B
Here’s some random junk
that’s left over but definitely worth showing:
