The demonstration
below shows the gamma induced errors for UnSharpMask in various processing
gamma spaces. The original in 16-bit mode is
linear 12-bit acquire using Canon eos*dcs 3 digital camera into Photoshop
15-bit space and then cropped.
The filter that
is used in this demonstration is UnSharpMask at Amount=400, Radius=0.5
and Threshold=2. Images were processed as follows:
|
Editing
space
|
Step
1: convert to editing by applying inverse gamma |
Step
2: convert to |
Step
3 |
Step
4: convert to viewing gamma space (2.2) by applying inverse gamma |
| 1.0 |
none |
8-bit/color |
Apply
USM |
2.2 |
| 1.2 |
1.2 |
8-bit/color |
Apply
USM |
1.83 |
| 1.4 |
1.4 |
8-bit/color |
Apply
USM |
1.57 |
| 1.6 |
1.6 |
8-bit/color |
Apply
USM |
1.37 |
| 1.8 |
1.8 |
8-bit/color |
Apply
USM |
1.22 |
| 2.0 |
2.0
|
8-bit/color |
Apply
USM |
1.1 |
| 2.2 |
2.2 |
8-bit/color |
Apply
USM |
none |
In step 5 the images
were saved as Jpeg at quality factor 4.
The overall gamma
is gamma in step 1 multiplied with gamma in step 4, that
is inverse gamma 2.2 for all the images.
Notice how your
vision (as the person who is editing the image) adapts to the quality
when you go slowly from linear editing down to gamma space 2.2 editing.
Then toggle
between linear editing and some of the gamma space processed images.
This is the difference in the first impression, when the image
(one of them) is presented to a viewer for the first time. Pay attention
e.g. to the alien appearance of the eye area and to the white gauze
around numbers in gamma space processed images.