Extensions - LMIP and Depth-Shaded MIP

Figure 4.8: a) MIP of an angiography with depth shading. b) LMIP of a stenosis of the aorta

(a)	(b)

The algorithm for MIP of value-sorted voxels can be easily extended to generate depth-shaded images (figure 4.8a). Three depth-templates are generated to calculate the depth of each voxel's projection from it's coordinates by look-up. The intensity value of each voxel is modulated by it's depth and written into the base plane image only if the value of the pixel at this position is smaller than the modulated value. As the depicted maximum values do not directly correspond to data values, applying this method to preprocessed volume data may produce results slightly differing from projecting non-preprocessed data.

The second extension of the algorithm is capable of generating LMIP images providing the possibility to interactively adjust the threshold parameter. For generating a MIP image, the order of examining samples along a ray is not relevant. Straight-forward LMIP requires the samples to be processed front to back along the viewing ray in order to find the first local maximum. Using templates for voxel-depth calculation and two z-buffers per pixel allows to extract the closest local maximum from samples arriving in arbitrary order. The z[pix] buffer stores the depth of the currently visible voxel, while zb[pix] stores a ``back-clipping'' distance for each pixel which is used to skip voxels belonging to maxima behind the currently closest one. As the voxels are processed in order of ascending data value, all voxels below the LMIP-threshold can be first projected using the simple and fast MIP algorithm without z-calculation. Among all voxels above the threshold which are processed later, closest local maxima have to be found. At the beginning, z[] is initialized to contain the maximum possible distance. With z and v containing the depth and value of a projected voxel, the closest local maximum along a ray is found by

if (z<z[pix])
  screen[pix]=v; zb[pix]=z[pix]; z[pix]=z;
else if (z<zb[pix])
  screen[pix]=v; z[pix]=z;

The first condition detects voxels closer than the currently displayed voxel. As they have at least the same intensity as the currently displayed one, they are entered into the screen and z[] buffers, the back clipping plane is moved forward to the previous z[] position (figure 4.9a). The second condition takes care of new voxels which are placed between the currently displayed voxel and the back clipping distance and are at least of the same intensity as the current one. In this case the screen and z[] are set to the new voxel. Voxels behind the back clipping distance are ignored (figure 4.9b). As the value at zb[] is always older than the value at z[] it is also smaller due to the ascending processing order of the voxels. Thus, any voxel which arrives later during processing and is located behind zb[] can be ignored , as it does definitely not belong to the closest maximum.

mailto:mroz@cg.tuwien.ac.at.