feat(ColorMaps): Add grayscale sigmoid colormap based on DICOM standard #3365
Conversation
Add a non-linear sigmoid grayscale colormap as an alternative to linear grayscale. This mapping is described in the DICOM standard in section C.11.2.1.3.1 SIGMOID Function and equation C.11-1. We use a normalized window_width of [-1, 1] and window_center = 0. The colormap is generated using a 16-point sampling of this function.
jadh4v
force-pushed
the
ENH-add-sigmoid-interpolation-to-transfer-functions
branch
from
18b448d to
aed71ec
Compare
|
cc: @finetjul please also review |
| -1, | ||
| 0.01798620996209156, | ||
| 0.01798620996209156, | ||
| 0.01798620996209156, |
There was a problem hiding this comment.
Good point. I am using the equation specified by the DICOM standard:
1 / (1 + e^(4x))
The multiplying factor of 4 could be increased to make the ends closer to 0 and 1 but they are asymptotic, so they only converge to those values. I think the purpose of the sigmoid is to increase the gradient at the center of the scalar range (window). And it is probably ok to sacrifice some color range at the ends.
| 1, | ||
| 0.9820137900379085, | ||
| 0.9820137900379085, | ||
| 0.9820137900379085 |
| "RGBPoints" : [ 0, 0, 0, 0, 1, 1, 1, 1 ] | ||
| }, | ||
| { | ||
| "ColorSpace" : "RGB", |
There was a problem hiding this comment.
why not rely on midpoint+sharpness ?
https://github.com/Kitware/vtk-js/blob/master/Sources/Rendering/Core/ColorTransferFunction/index.js#L721-L725
There was a problem hiding this comment.
Maybe add support for a "Nodes" key/value in addition to "RGBPoints".
See applyColorMap in ColorTransferFunction/index.js
2 participants
Context
Add a non-linear sigmoid grayscale colormap as an alternative to linear grayscale. This mapping is described in the DICOM standard in section C.11.2.1.3.1 SIGMOID Function and equation C.11-1. We use a normalized window_width of [-1, 1] and window_center = 0. The colormap is generated using a 16-point sampling of this function.
See: https://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_c.11.2.html
Results
A comparison of linear vs sigmoid grayscale using the ImageViewer example.
The sampled 16-point sigmoid function plotted through matplotlib:
Changes
PR and Code Checklist
npm run reformatto have correctly formatted codeTesting