Barry “BoB” Hon’s Project 4 Results

Project 4’s auto stitching involved steps:

1) Harris edge detection

2) ANMS choosing of points

3) Point matching

4) Ransack

5) Stitching

Pool revisited

Step 1

This first set shows Harris edge detection. There are many points that need to be removed

Step 2

Using ANMS we remove points but attempt to keep the points evenly distributed

Step 3

After this we match points only matching those that have a threshold of NN-1/NN-2 > 0.5 (Nearest Neighbor)

Step 4

Finally we run ransack to pick points that closely match each other and get rid of outliers.

Step 5

We then stitch the image with function from Project 3

Here’s another stitch:

The points chosen after harris, anms, matching and ransac:

Stitched:

Here’s a sample from the data set provided on the BB. There is a line in the middle of the picture because of the way I did merging (with use of Distance maps) in Project 3, but the stitching and point matching itself is correct.