DIRECT VOLUME RENDERING Flashcards
What is volume rendering
Concerned with rendering the inside of things
Both Amorphous (clearly defined boundaries) or not
its not about surfaces
Common things modelled that dont have boundaries or surfaces
weather fronts
heat
sea
What is a core issue of data points in volume rendering
Knowing how they are classified: core piece or external
What are the data point units in a CAT scanner
radiodensity values measured in Hounsfield units
-1000 Air
-600 lung tissue
0 water
40 muscle
350 bones
1800 skull bone
(don’t need to remember these)
What is the transfer function
Transforming data values to visualisable data
What is a voxel
a single sample, or data point, on a regularly spaced, three-dimensional grid
We can make voxels smaller and sample at a higher frequency to get a more definitive image
Raycasting for volume rendering
A more sophisticated method
Fire a ray from the viewpoint, through the pixel in our viewplane and into the volume
when the ray enters the volume, we march along in tiny steps and measure the colour and opacity associated with the voxel we happen to be in
We store this information
Repeat this process, casting many rays
What is additive reprojection
The technique being used in raycasting for volume rendering
flat image of a 3d object, pixels change colour depending on what the ray has passed through
Create a surface using surface normals
We create a fake surface normal by detecting changes in the voxel in three different directions (x,y,z)
The directions do not actually mean anything, the normal is symbolic - but still helps us create the surface
We can plug this faux surface normal vector into local illumination equations to create surface shading
The fundamentals of direct volume rendering
Essentially we are just picking colour and opacity values - we are not calling anything as directly transparent or naming colours too crudely
All our data should pass through this direct volume rendering process
Really important when say looking for a large clump of cancerous cells - but we are still relying on our brain to do a lot of the visualisation
Cons of direct volume rendering
If we want to move the viewpoint, we have to re do the whole ray casting step - computationally long
Good quality costs at significant computational costs
- DVR is not supported end to end by computer hardware