Description
This assignment deals with cubic b-spline filtering applied to the image produced by the equation:
L( x, y ) = 0.5 * ( 1 + sin( (x^2 + y^2) / 100 ) ).
In addition, jittered sampling without filtering and multi-jittered sampling, both with and without filtering, have been included for comparison.
Base Image
This is the graph of L( x, y ) with a single sample centered in each pixel.
Filtered Images
Below are graphs of L( x, y ) using jittered and multi-jitter sampling, and tent, cubic b-spline, or no filtering.
| Sampling Method | 1 sample / pixel | 16 samples / pixel | 64 samples / pixel | 256 samples / pixel | 900 samples / pixel |
|
Jittered |
 |
 |
 |
 |
 |
| Multi-Jittered |  |
 |
 |
 |
 |
| Tent Filter with Jittering |  |
 |
 |
 |
 |
| Tent Filter with Multi-Jittering |  |
 |
 |
 |
 |
|
Cubic B-Spline Filter with Jittering |
 |
 |
 |
 |
 |
|
Cubic B-Spline Filter withMulti-Jittering |
 |
 |
 |
 |
 |
All of the sampling methods above are significantly better than using a single, centered sample for each pixel, assuming enough samples to remove most noise. The cubic b-spline filter clearly performs better than the tent filter. Multi-jittering seems to marginally decrease noise, but does not significantly improve the quality of the image.