Image Mosaicing
Image stiching and panorama generation.
The purpose of this project was to find corner features in multiple images and to align the images
in a mosaic by estimating a homography between corresponding features
The basic outline of this project followed the below steps.
- Converting the images to grayscale and reading it.
- Apply Harris corner detection algorithm and compute Harris R function over the images.
- Find correspondences between the obtained corner points of two images by performing Normalised Cross Correlation (NCC).
\begin{equation}
NCC= \frac{\sum_{i=1}^{25}x(i)y(i)}{\sqrt{(\sum_{i=1}^{25}x^2\sum_{i=1}^{25}y^2)}}
\end{equation} - stimate the homography using the above correspondences by utilising RANSAC.
- Rectify the images according to the homography matrix obtained from above.
Homography Matrix - Just like fundamental matrix in previous project, the Homography matrix is a matrix that states that any two images with same planar surface in space are related by a transformation matrix with 8 degrees of freedom. This matrix is mainly ussed for immage rectification. \begin{equation} \begin{bmatrix} lx’\\ ly’\\ 1 \end{bmatrix} = \begin{bmatrix} h11 & h12 & h13\\ h21 & h22 & h23\\ h31 & h32 & h33\\ \end{bmatrix} \begin{bmatrix} rx\\ ry\\ 1 \end{bmatrix} \end{equation}
where l is left image pixel and r is right image pixel.
