Applications of optical techniques in interconnections, image/video compression, and computational geometry

by Yoshida, Akitoshi, Ph.D., Duke University, 1994, 164 pages; AAT 9424616



Abstract (Summary)

This thesis presents new applications of emerging optical computing techniques to important computer science problems. Three areas are identified as target application areas because of their importance and the possible benefits of optical computing techniques. These three areas are processor/memory inter-connections, image/video compression, and computational geometry. New approaches are designed and investigated in these areas, and the motivations for using optical techniques to overcome the limitations of traditional approaches are discussed.

First, the problem of addressing a large number of processor/memory cells is investigated. The problem is very important for message routing and memory addressing in highly parallel computers. Approaches are divided into two different methods: one using matrix-multiplications and the other using deflectors. The scalability of these schemes is analyzed from the basic properties of the optical systems. The results are found to be advantageous over conventional approaches.

Second, image/video compression techniques are investigated. Conventional implementations require a large number of computations. A cosine transform system is designed from a simple optical system. Elaborate vector quantization systems are designed from optical systems with photorefractive crystals. Block matching systems for video compression are designed. These systems require only constant time; thus for time critical applications, they are advantageous.

Third, computational geometry problems are discussed. Basic geometric transformations using computer generated holograms are demonstrated by simulation. In particular, the discrete cell multiplexed holograms are used to transform points. They are suited for computational geometry problems and give various single-step algorithms.

This thesis will not only introduce new ideas in these application areas but also steer the field of optical computing in new directions through these applications.

Indexing (document details)

Advisor:

Reif, John H.

School:

Duke University

School Location:

United States -- North Carolina

Source:

DAI-B 55/04, p. 1526, Oct 1994

Source type:

Dissertation

Subjects:

Computer scienceOptics

Publication Number:

AAT 9424616