Introduction

In this project, I produce a "morph" animation of my face into someone else's face, compute the mean of a population of faces, and extrapolate from a population mean to create a caricature of myself and more.

Part 1. Defining Correspondences

First, I cropped and rescaled an image of myself and an image of Tom Cruise to have the same dimensions so that our faces have a similar size and aspect ratio. I then used the spec-provided online tool to select 24 correspondence points for each of the two images (including the four corners). Finally, I generated a Delaunay triangulation at the midway points of the two sets of correspondence points.

Results

Part 2. Computing the "Mid-way Face"

In this section, I compute the "mid-way face" by averaging the geometry and appearance of two input images. The steps to achieve this are outlined below:

Methodology Overview

1. Compute Average Shape:

   The average shape is computed by taking the midpoint of each corresponding keypoint between the two faces. This creates a set of midway key points , representing the geometric average of both faces.

2. Delaunay Triangulation:

   Next, I apply a Delaunay triangulation to the midway keypoints. This divides the face into triangles, used for affine transformations of both input faces.

3. Affine Transformation Calculation:

   For each triangle from both original faces, an affine transformation matrix is computed using the computeAffine() function. This matrix maps the triangles from their original positions to the corresponding triangle in the average face.

4. Warping Faces into the Average Shape:

   Each face is warped into the average shape by applying the affine transformation for each triangle. A polygon mask is generated using polygon() to isolate the pixels in each triangle, and the transformation is applied to warp the original image pixels to the average shape.

5. Combine Warped Faces:

   Once both faces have been warped into the average geometry, I compute the final "mid-way face" by averaging the pixel values of the two warped images. This gives us a face with the average geometry and a blend of the colors from both input images.

Part 3. The Morph Sequence

In this part, I create a morph sequence between two images (A and B) using the key points and the Delaunay triangulation structure from Part 2. The morph sequence smoothly transitions between the two images by gradually warping and cross-dissolving them over multiple frames.

Methodology Overview

1. Warp and Dissolve:

   I use the key points from both images to interpolate between them. The warping is controlled by warp_frac, which transitions the shape between the two images, while dissolve_frac controls the blending of the colors from both images.

2. Morph Function:

   The morph() function takes two images, their key points, and the Delaunay triangulation to warp each face into an intermediate shape. For each frame, warp_frac and dissolve_frac are adjusted to gradually transition between the original and final images.

3. Create Sequence

   The morph sequence is generated by creating multiple frames, each corresponding to a different value of warp_fraction and dissolve_fraction in the range [0,1].