On some pel level research problems

摘要

The note lists some unsolved research problems concerning line (or edge) drawing algorithms for raster displays which seem to be still outstanding in 1997. Work in progress includes the so called "pixel perfect" speeded up versions of Bresenham's (1965) line drawing algorithm designed to exploit the parallel capabilities of modern display processors. Where should the build up of multi-pixel segments be terminated for optimum performance? The representation of circles, conic sections and other curves, by short, straight line segments or "otherwise" by direct generation at the pixel level: what is the "best" chain code definition? Is it "better" to use four-way or eight-way connected outlines? Can one use "strength reduction" to improve the performance of the algorithms when implemented with modern parallel hardware? How can one best convert between the "real" (floating point) coordinates of a real world model and the integer working required at the pixel level? Finally, how can one best use grey scale when it is available to soften the jaggies which so often spoil the appearance of fount outlines and other objects?.