Peter C. Wayner

Finding convex edge groupings in an image

In an image, there are groups of intensity edges that are likely to have resulted from the same convex object in a scene. A new method for identifying such groups is described here. Groups of edges that form a convex polygonal chain, such as a convex polygon or a spiral, are extracted from a set of image edge fragments. A key property of the method is that its output is no more complex than the original image. The method uses a triangulation of the linear edge segments in an image to define a local neighborhood that is scale invariant. From this local neighborhood a local convexity graph is constructed; this encodes which neighboring image edges could be part of a convex group. A path in the graph corresponds to a convex polygonal chain in the image, with a cyclic path corresponding to a polygon. We have implemented the method and found that it is efficient in practice as well as in theory. Examples are presented to illustrate that the technique finds intuitively salient groups, including for images of cluttered scenes.

Efficiently using invariant theory for model-based matching

A method is presented for efficiently maintaining and searching a database of three-dimensional models so they can be reliably recognized from arbitrary two-dimensional projections in the presence of noise and occlusion. The core of the process is the topologically defined network of invariants which breaks three-dimensional models down into small, local groups of features and indexes these groups using functions that are invariant under translation, rotation, scaling, and orthographic projection. The network encodes the geometrical relationships between these groups so that grouping information can be used to increase the speed of matching.<<ETX>>

Content-Addressable Search Engines and DES-like Systems

A very simple parallel architecture using a modified version of content-addressable memory (CAM) can be used to cheaply and efficiently encipher and decipher data with DES-like systems. This paper will describe how to implement DES on these modified content-addressable memories at speeds approaching some of the better specialized hardware. This implementation is often much more attractive for system designers because the CAM can be reprogrammed to encrypt the data with other DES-like systems such as Khufu or perform system tasks like data compression or graphics.The CAM memory architecture is also easily extendable to build a large scale engine for exhaustively searching the entire keyspace. This paper estimates that it will be possible to build a machine to test 255 keys of DES in one day for

Boxweb: A Structured Outline Program for Writers

30 million. This design is much less hypothetical than some of the others, in the literature because it is based upon hardware that will be available off-the-shelf in the late end of 1992. The architecture of this key search machine is much more attractive to an attacker because it is easily reprogrammable to handle modified DES-like algorithms such as the UNIX password system or Khufu.

Method for determining boundaries of words in text

Boxweb is a high-level language for creating outlines. It forces the writer to list both the ideas in an essay and more importantly the way the ideas interact. These lists are compiled by the computer into a traditional, hierarchical outline. The structure of Boxweb is designed to force the writer to think of an essay as not just a set of facts and notions, but as a web of connections between ideas. While the basic structure of the language is adapted from computer languages, the syntax is specifically designed to be very flexible and appear like natural English. The paper will describe the structure of the language, some results from testing it on writers and finally, the algorithms used by the compiler. It will compare the language to some related work done at Xerox P ARC, which is primarily visual.