Please click on the thumbnails to see the full-resolution images and use the back button to return to this page.
We provide a comparison of our algorithm with three state of the art technics for strong detail enhancement. For the domain transform and guided filter, we used the parameters suggested in the papers and a 5x enhancement. For [Fattal 2007] multiscale method, we used the parameters of the paper and multiplied the output by 2 to have an exposure more similar to the other methods. For adaptativ manifolds, we used the parameters suggested in the paper on the grayscale image to avoid strange color changes.
On the small versions of the images, it sometimes looks like the images of the other algorithms have more details, but you can click on them to see that the impression is mostly given by different constrast and that our full resolution images have more details.
On the images with strong edges, halos appear in the other algorithm, but not or much less with the Local Laplacian filter
| Original | our algorithm with sigma=0.5 and fact=6 | Domain transform [Gastal and Olivera 2011] | our algorithm with sigma=0.1 and fact=5 | Guided filter [He et al. 2010] | Adaptativ manifolds [Gastal and Olivera 2012] | Unnormalized bilateral filter (our algorithm with only one scale) sigma=0.05 and fact=9 | Multiscale [Fattal et al. 2007] |
|
|
|
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|