CS194 Project 5

Image Warping and Mosaicing

Xuxin Cheng CS194-agv

Part 1 Image Warping and Mosaicing

1.1 Recovering Homogprahy

A Homography matrix is a 3x3 matrix with 8 degree of freedom. To solve this matrix, we need at least 4 pairs of corresponding points. If we have an overestimated linear system that has more than 8 equations, we can use Least Squares to solve this problem.

$\begin{bmatrix}wx_1\\wx_2\\w\end{bmatrix}=\begin{bmatrix}h11&h12&h13\\h21&h22&h23\\h31&h32&1\end{bmatrix}\begin{bmatrix}x\\y\\1\end{bmatrix}$

deliminate $w$ and combine other point pairs we can derive: $Ah=b$, where h contains the 8 variables in homography matrix. Solving this least square problem we will have H, which is the homography matrix.

1.2 Image Rectification

original image	rectified image

1.3 Blend the images into a mosaic

image1	image2

Blended image