lzyhm wrote:
> stefanbanev@yahoo.com wrote:
> > lzyhm wrote:
> > > stefanbanev@yahoo.com wrote:
> > > >
> > > > One of mentioned feature is "Surface Generation". The impulse
> > > > opacity transfer function (step like) should provide iso-surface so,
> > > > why this very particular case of volume rendering is distinguished in
> > > > separate nomination?
> > > >
> > > If you define left eye ROI slice by slice, tviewer can generate the
> > > surface by ROIs even the slices have different spacing.
> >
> > The original data provides just set of interpolation grids thus, what
> > so special about a non-regular spacing. The iso-surface of volume is
> > still perfectly defined from point of view of interpolator (tri-linear
> > or higher) so, impulse transfer function must show the correspondent
> > iso-surface of scalar field seeing by interpolator. The quality of
> > iso-surface obtained through step-TF is heavily depends on
> > super-sampling level and order of interpolation (tri-linear or higher).
> > To get comparatively accurate mesh-based iso-surface you will need at
> > least 8 polygons per voxel what actually excessively far more expensive
> > than super-sampling ray-casting approach. Definitely the polygonal
> > substitution is a viable option for many applications where a draft
> > approximation is acceptable.
>
> The generation of iso-surface based on marhing-cubs commonly. the
> marhing-cubs'input is auto-segmentation result by impulse transfer
> function. The speed is slow, the more importance is the
> auto-segmentation result has many other uncorrect infomation, many
> arithmetic must be distribute to remove these uncorrect segmentation.
> and the mesh-based iso-surface is diffcult for advance computation such
> statistic or DVH evaluation etc. The define ROIs by hand still is used
> in many situation.

Marching cube may provide very rough iso-surface representation,
however for many applications it may sufficient.


Thanks,
--sb