Lighting artifacts due to near-zero gradient magnitude

[mlavik1]

In homogeneous areas of the dataset, the gradient magnitude will be near zero and using the gradients as normals for lighting calculations may cause visual artifacts.

Possible solution:
Modify lighting calculations to only shade areas with a high gradient magnitude, as suggested here.

Question: How do we define "significant" gradient magnitude, when different datasets will have varying densities and sizes? As a percentage of the maximum gradient magnitude in the dataset, or as hounsfield values?

[mlavik1]

See #208