Based on this blog post.
Chose algorithm :
- Naïve
- Two pass
- Two pass with linear sampling
and filter size :
- LITTLE_KERNEL (7)
- MEDIUM_KERNEL (35)
- BIG_KERNEL (127)
| LITTLE_KERNEL 3 sigma | MEDIUM_KERNEL 10 sigma | BIG_KERNEL 25 sigma |
|---|---|---|
 |
 |
 |
At this moment kawase blur have not param 5 pass with [0, 1, 2, 2, 3] offset ...