Project 4 Part 1: Image warping and mosaicing!
CS194-26 Fall 2021 | Rio Hayakawa
Introduction
In this project, we will be implementing an auto panoramic mosaic stitcher using multiple images shot from the same location. In this part, we define corresponding points in each of the images and calculate a homography matrix to warp the images to one another to align and stich them.
Shoot and Digitize Images
Images for stitching:
Wurster Left |
Wurster Right |
|
|
Images for rectifying:
Recover Homographies
In order to get the transformation matrix between the images, we must recover the homography matrix H as below.
We can do this by solving a linear system set up using the corresponding points between the images as below.
Warp the Images
Using the parameters of homography, we warp the left image according to the corresponding points of the right image.
Before Warp |
After Warp |
|
|
Image Rectification
We can demonstrate that the warping is working by warping an object that we know is a rectangle and assigning the corresponding points to be such.
Before Rectifying |
After Rectifying |
|
|
Before Rectifying |
After Rectifying |
|
|
Blending the images into a mosaic
Conclusion
The coolest thing I learned was that we can make out patterns and text from an image of a rectangular object just by using a homography and rectification. It’s fun that we can stretch out a certain perspective of an image to make it look like we are looking at the object straight on.
Project 4 Part 1: Image warping and mosaicing!
CS194-26 Fall 2021 | Rio Hayakawa
Introduction
In this project, we will be implementing an auto panoramic mosaic stitcher using multiple images shot from the same location. In this part, we define corresponding points in each of the images and calculate a homography matrix to warp the images to one another to align and stich them.
Shoot and Digitize Images
Images for stitching:
Images for rectifying:
Recover Homographies
In order to get the transformation matrix between the images, we must recover the homography matrix H as below.
We can do this by solving a linear system set up using the corresponding points between the images as below.
Warp the Images
Using the parameters of homography, we warp the left image according to the corresponding points of the right image.
Image Rectification
We can demonstrate that the warping is working by warping an object that we know is a rectangle and assigning the corresponding points to be such.
Blending the images into a mosaic
Conclusion
The coolest thing I learned was that we can make out patterns and text from an image of a rectangular object just by using a homography and rectification. It’s fun that we can stretch out a certain perspective of an image to make it look like we are looking at the object straight on.