Computational Photography – Project 3
Face Morphing
·
Corresponding
points are found in both the start and finish images, which are in this case
two people’s faces
·
The
average structure is found by taking the mean of the two sets of points, a Delaunay
triangulation is found using these points
·
For
each frame of the morph an intermediate set of points is found by taking a
weighted average of the two original sets of points, the weights depend on the
frame
·
An
affine transform is found for each triangle in the intermediate points to both
the start and end sets of points
·
Each
pixel in the intermediate image is transformed to both the start and end images,
this will return non-integer values so interpolation is used
·
The
interpolated pixel values are used to make a warped version of the start and
end images
·
The
two warped images are combined using a weighted average with the same weights
as above
Morphing from 19_hugh to 20_minjae
This gif has only 10 frames. The full
61 frames can be found here.
·
Corresponding
points are found in the images from the faces of the whole class
·
The
average structure is found by taking the mean of all the sets of points, a Delaunay
triangulation is found using these points
·
Each
face is warped to the mean structure using the method described above for two faces
·
All
the warped images are averaged to give the mean face
My face warped to the mean geometry
To give the mean face an identity I
took the average of the mean face image and my face warped to the mean geometry
To touch up my face I first warped
the mean face image to my geometry then took the average of my face and the
warped mean face.
The mean face warped to my geometry
The original image of my face and
after averaging with the warped mean face