Cory Seager |
Mookie Betts |
Firstly, I selected 52 key points from seager and betts image. They were based on the features of faces. As well as features of the face, I selected four points from four corners of the image. Then, I computed mean of two key points and used them for Delaunay's triangulation.
Firstly, I computed the average shape by taking average of key points from two images. Then, I warped both images to the average shape. To compute affine matrix, I solved linear system. Then, for warping, I obtained pixels within triangle using polygon. Then, I did inverse warping to obtain warped image. Once I got two warped images, I averaged the color of two images for midway face.
Original Seager |
Mid Way Face |
Original Betts |
For morphing, I did similar thing as part 2. I morped each image 45 times since we needed 45 frames. There were two fraction parameter, warp fraction and dissolve fraction. For each morphing, fraction values were different. I warped two images to the shape that was determined by warp fraction. Then, I combined these two warped images according to dissolve fraction.
In this part, I used Danes image (240 totals). 58 key points were given, and I added four corner points. Firstly, I computed the average shape of these points of 240 images. Then, I warped all 240 images to this average shape. Below, I provided some examples of warped images to their average shape.
After obtaining 240 warped images, I took an average of them to get an average face.
For the last step, I warped my face to average face geometry, and then warped average face to my face geometry.
My face
My face to average face |
Average face to my face |
In this part, I first computed caricatures point which are extrapolated of points from my face. I used "extrapolated point = mean_point + alpha(my_point - mean_point)". Then, I warped points from my face to extrapolated points. I found several caticatures for several different alpha value.
alpha = 0.4
alpha = 0.7
alpha = 1 (same as original image)
alpha = 1.3
alpha = 1.6
In this part, I used my face and one woman's face from Danes. I selected key points from two images, then I took a mean of key points from two images. Then, I warped my image (morphing the shape) and woman's image to mean shape. Then, I dissolved these two warped images to obtain morphed image (morphing the shape and appearance.
Morphing the shape
Morphing the appearnace
Morphing the shape and appearance
Seager:https://www.mlb.com/player/corey-seager-608369
Betts:https://www.mlb.com/player/mookie-betts-605141