Experiment 1
Region-based matching
| Parameters | |
| Score | SAD |
| Template size | 30 |
| Match window width | 9 |
| Match window height | 5 |
| Results | |
| Calculated disparity map |  |
| Difference to true disparity map |  |
| Anaglyph after warping |  |
The assignment can be found here. The experiment results are below.
| Parameters | |
| Score | SAD |
| Template size | 30 |
| Match window width | 9 |
| Match window height | 5 |
| Results | |
| Calculated disparity map | |
| Difference to true disparity map | |
| Anaglyph after warping | |
| Parameters | |
| Score | SAD |
| Template size | 30 |
| Match window width | 9 |
| Match window height | 5 |
| Results | |
| Calculated disparity map |  |
| Difference to true disparity map |  |
| Anaglyph after warping |  |
| Parameters for level 4 | |
| Method | Feature-based matching |
| Score | SAD |
| Template size | 4 |
| Match window width | 3 |
| Match window height | 3 |
| Calculated disparity map |  |
| Parameters for level 3 | |
| Method | Region-based matching |
| Score | SAD |
| Template size | 10 |
| Match window width | 5 |
| Match window height | 5 |
| Calculated disparity map |  |
| Parameters for level 2 | |
| Method | Region-based matching |
| Score | SAD |
| Template size | 15 |
| Match window width | 7 |
| Match window height | 5 |
| Calculated disparity map |  |
| Parameters for level 1 | |
| Method | Region-based matching |
| Score | SAD |
| Template size | 20 |
| Match window width | 9 |
| Match window height | 5 |
| Results | |
| Calculated disparity map |  |
| Difference to true disparity map |  |
| Anaglyph after warping |  |
The multi-resolution matching method gives the best results.
In particular, multi-resolution matching is more robust against bad parameters, since it builds on disparity values from the previous level, which are usually less noisy.