Texture-map computations can be made tractable through use of precalculated tables which allow computational costs independent of the texture density. The first example of this technique, the “mip” map, uses a set of tables containing successively lower-resolution representations filtered down from the discrete texture function. An alternative method using a single table of values representing the integral over the texture function rather than the function itself may yield superior results at similar cost. The necessary algorithms to support the new technique are explained. Finally, the cost and performance of the new technique is compared to previous techniques.

A supervisory process is used to distribute picture-generation tasks to heterogeneous subprocesses. Significant advantages accrue by tailoring the subprocesses to their tasks. In particular, scan conversion algorithms tailored to different surface types may be used in the same image, a changing mixture of processors is possible, and, by multiprogramming, a single processor may be used more effectively. A two-level shape data structure supports this execution environment, allowing top-level priority decisions which avoid comparisons between surface elements from non-interfering objects during image construction.

A video-resolution frame buffer system with 32 bits per pixel is described. The system includes, in addition to standard features for limited zoom and pan, an arithmetic unit at the update port which allows local computation of many frequently-used pixel-level functions combining stored pixel values with incoming pixel values. In addition to the standard arithmetic and logical functions there are functions for sum to maximum pixel value and difference to minimum pixel value. Comparisons between incoming and stored data are used to implement conditional writes based on depth values for depth-buffer algorithms. Update and refresh ports are designed for a wide range of flexibility allowing simultaneous use by separate tasks and various functional rearrangements of the 32-bit pixel words. The memory architecture, refresh and update ports are described. Examples of widely divergent modes of operation are provided.

Decreasing memory costs will soon allow grayscale displays in low-cost raster graphic terminals. Subtle shadings can be used to provide improvements in line quality and character flexibility which could allow raster displays to compete on better terms with the more expensive calligraphic displays. Algorithms for achieving smooth vectors and rotatable dot matrix characters are outlined and scan conversion is discussed. A discussion of the relation between image quality and number and distribution of gray levels follows, with concluding remarks on costs and other practical matters.

Shadows are advocated for improved comprehension and enhanced realism in computer-synthesized images. A classification of shadow algorithms delineates three approaches: shadow computation during scanout; division of object surfaces into shadowed and unshadowed areas prior to removal of hidden surfaces; and inclusion of shadow volumes in the object data. The classes are related to existing shadow algorithms and implementations within each class are sketched. A brief comparison of the three approaches suggests that the last approach has the most appealing characteristics.