Table of Contents
HW 4: Convolution
Source Code
- AFS Link to Executable: /afs/andrew.cmu.edu/usr23/nhergert/gitrepo/16-311GeneralRobotics/hw4convolution/hw4
- Source Code: hw4.c
Results
| 3×3 Mask | 5×5 Mask | 9×9 Mask | |
|---|---|---|---|
| White = 1 Gray = 0 Black = -1 |  |  |  |
| Original | Mask 1 | Mask 2 | Mask 3 |
|---|---|---|---|
 |  |  |  |
 |  |  |  |
Analysis
- What do these masks do to the input image?
Each mask performs a differentiation on the pixels surrounding it and generally highlights the “edges” of the original image.
- What are the advantages of using a small mask? Why would you want to use a large mask?
Using a small mask allows for very high detail, whereas a larger mask blurs over some of the noise in the image because it samples from a larger area but also smooths over sharp edges.
- Suppose you wanted to identify the horizon in images at sea. What mask might you use to do this?
The masks used in this experiment are good only for vertically-oriented edges, whereas these masks flipped 90° would perform much better for a horizon at sea and the horizontal edges in these images.