CS194 Project 6A: Image Warping and Mosaicing

Kristine Lin 25523003

Image Rectification

To create rectified images, I would choose 12 points and map them to be a grid-like form. Since the images that I chose in this part were all of tiles, this was able to warp the images into flattened form where the tiles are viewed as rectangles/quares. Using 8+ points enabled me to solve for a system of linear equations (using least squares) and form the following images.

Tiles on the ground

tiled_floor

Tiles on the ground rectified

tiled_rectified

Tiles on the ground

tiled_room2

Tiles on the ground rectified

tiled_warped2

Mosaics

To create the mosaics, I would first take two images and define the points of correspondance (same as warping). Again, I solved for H and warped the second image to the first image. To stitch the two images together, I blended the two images using weighted sum. I feathered out the image the further the image got from the center and summed them together.

Original Images #1

im1 im2

Mosaic #1

blended_image1

Original Images #2

im4 im5

Mosaic #2

blended_image2

Original Images #3

im6 im7

Mosaic #3

blended_image3

Summary

I learned how to find the homography matrix and how much work goes into the panorama photos. It's pretty cool how much you can manipulate 2D photos to provide a 3D viewing experience.