Anusha Syed CS194-AAU, Project 4

The Mid-Way Face and Morph Sequence

The strategy I used was the same as we discussed in lecture. The first thing I did was compute a correspondence of points between the two images, for example, the eye in image a and the eye in image b. After obtaining a set of several correspondence points, I computed the average point of the points by averaging each point in image a with the corresponding point in image b. I then morphed each to this average shape, known as the midway face. In order to do this, I computed a Delaunay triangulation of the average shape points and applied it to the original image to find the points that corresonded to each triangle of the triangulation. Then, for each point that formed the correspondent triangle, I performed an affine transform between the points and the triangle from the triangluation. The resultant pixel for the destination image was given by multiplying the inverse of the affine matrix solved for in the last step. Although this may have been the correct approach, there was unforunately, and quite obviously, had a significant bug in my code that resulted in a large triangular seam across my image. Even a couple all nighters later, I was not able to locate the bug unfortunately, but it is most likely was result of incorrecltly computing the relationship between the corresponding triangles.

The Mean face of a population

I computez the mean face of the male images from the FEI face database.

Caricatures: Extrapolating from the Mean

Bells and Whistles: Warp between friends