Part 1. Defining Correspondences
In this part of the process, I define corresponding points on two images of faces to facilitate morphing between them. I manually select pairs of keypoints on both images, ensuring consistent labeling between the two faces for accurate morphing. Using a tool created by a former student, found here, I plot and save these points for future use. I compute a Delaunay triangulation based on the average shape of the two sets of points, creating a well-defined mesh that avoids overly skinny triangles. This triangulation is applied uniformly across the morph.
Part 2. Computing the "Mid-way Face"
In Part 2, I compute the "mid-way face" between images of me and Christian Bale. I calculate the average shape by averaging the keypoints from both faces. Then, I warp both faces into this average shape using affine transformations for each triangle in the triangulation. The warped images are combined by averaging the pixel values, resulting in the mid-way face.
Part 3. The Morph Sequence
I created a morph sequence transitioning between two images using point correspondences and a triangulation structure. The sequence is generated by gradually changing the shape (warp_frac) and blending colors (dissolve_frac) between the two images. Below is a set of images from the morphing sequence, as well as an animated GIF that visually demonstrates the smooth transition from my face to Christian Bale's.
Part 4. The "Mean Face" of a Population
For this part, I used the Danes dataset to compute the average face shape of a population. I selected keypoints across the dataset, morphed each face into the average shape, and then computed the average face. I also warped my face to match the average geometry and warped the average face to match mine.
Part 5. Caricatures: Extrapolating from the Mean
I created a caricature of my face by extrapolating from the population mean face I computed in the previous step. To do this, I exaggerated the differences between my facial features and the average face. By scaling the difference between my face's keypoints and the population mean, I pushed my facial features further away from the mean, which accentuates my unique characteristics.
Bells and Whistles
In my project, I performed facial morphing to change my appearance by using an average image of a Chinese man as a reference.
Average Chinese Man
Me
Shape Only
Appearance Only
Shape and Appearance
