Project 4

Felix Juefei Xu

juefeix@andrew.cmu.edu

Nov.11 2010

Image Warping and Mosaicing: (Part A)

In this project, two or more photographs are warped, registered and mosaiced.

Here are 2 input images:

The blue and red dots in both of the input images are the corresponding pairs formed in accordance with the following:

Once the corresponding points are acquired, I can compute the projection matrix H which is 3x3. And then warp image 1 to image 2 as following:

Here is the example of rectification of this warping image:

Here is the final blending result:

Feature Matching for Autostitching : (Part B)

In this part, I follow these steps for realising feature matching for autostitching:

1. deteting corner features in an image

2. exracting a feature descriptor for each feature point

3. matching these feature descriptors between two images

4. use RANSAC to find homography H

I start using Harris interest point detector to detect the interest points for my input images:

Here are the interest points detected:

For each interest point, an 8x8 patch centered at the interest points are extracted to be feature descriptor. In this simple case, we just use the direct pixel as features.

After feature descriptors are formed for each interest point, Enclidean distance is used for matching. Once we know which points have the highest 1NN/2NN ratio, we can use those points to find the homography H using RANSAC technique.

And the warping part is the same as Part A of this project.

END