In complex 3D scenes, important objects (objects of interest) are often
occluded by less important objects (secondary objects). The cutaway technique
addresses this problem by exposing objects of interest and removing secondary
objects that occlude them. The adaptive method presented by Burns and
Finkelstein[1] performs the cutaway action dynamically, depending on the view
vector (see Figure 1).
The screen-space image of the scene forms the basis of this method. For each
pixel, view-dependent depth values are compared.
At first, the occluding region which is going to be cut away has to be determined.
It is represented by the boundary surface between the occluding and the
non-occluding region which is called the cutaway surface. It is designed to
approximate the silhouette of the object of interest. For multiple objects, it
approximates the silhouette of the union of the objects. The cutaway surface is
used to decide whether a fragment is removed or not by determining if the
fragment lies before or behind that surface as seen from the viewer. Respective
depth values are calculated using an alternative version of the chamfer distance
transform. They are computed for a specified viewing angle which has to be
chosen accordingly to avoid artefacts:
Interior surfaces of secondary objects exposed by the cutaway procedure are called cut surfaces. Visual quality is improved by highlighting the contours of cut surfaces. Additionally, dashed ghost lines illustrate removed geometry.
The implementation is based on a windows applictation that is formed of following components:
The following screen shots show some results of the running application. The first session is from the scene Japanese Housing and shown in table 1, the second session is from the scene Villa Vanila and shown in table 2.