The main idea here is similar to the one in Weiler and Atherton [1], which operates by adding objects incrementally according to a depth order. At each stage in Weiler and Atherton’s algorithm, the newly added object is clipped against the current visible surfaces. The main difference here is that instead of adding one object at a time, the number of objects added in the next phase is dependent on the complexity of the contour of the currently visible surfaces. It is this technique that is responsible for the “output sensitive” complexity of the algorithm.
The authors first present their solution for triangular primitives and then extend it to other objects, such as polygons, disks, and balls. The complexity analysis for the extension to concave polygons using the polygon cutting theorem is not obvious and could have been addressed in greater detail. Moreover, a distinction needs to be made between tessellation edges and actual boundary edges in order to obtain the correct final result. The algorithm assumes that no cyclic overlaps (in depth) occur among the objects in the scene, and it is restricted to such situations. It would therefore be useful to compare the complexity of detection of cyclic overlaps to the complexity of computing visible surfaces. The authors point out that even image space techniques have problems with cyclic overlaps. The commonly used Z-buffer hidden surface algorithm does not have this problem.
Overall, the paper is clear. Also, as claimed in the title, the algorithm appears to be simple.