Abstract

Demands for fast execution of image processing are a driving force for today's computing market. Many image processing applications require intense numeric calculations to be done on large sets of data with minimal overhead time. To meet this challenge, several approaches have been used. Custom-designed hardware devices are very fast implementations used in many systems today. However, these devices are very expensive and inflexible. General purpose computers with enhanced multimedia instructions offer much greater flexibility but process data at a much slower rate than the custom-hardware devices. Digital signal processors (DSP's) and media processors, such as the MAP-CA created by Equator Technologies, Inc., may be an efficient alternative that provides a low-cost combination of speed and flexibility. Today, DSP's and media processors are commonly used in image and video encoding and decoding, including JPEG and MPEG processing techniques. Little work has been done to determine how well these processors can perform other image process ing techniques, specifically image rendering for printing. This project explores various image rendering algorithms and the performance achieved by running them on a me dia processor to determine if this type of processor is a viable competitor in the image rendering domain. Performance measurements obtained when implementing rendering algorithms on the MAP-CA show that a 4.1 speedup can be achieved with neighborhood-type processes, while point-type processes achieve an average speedup of 21.7 as compared to general purpose processor implementations.

Library of Congress Subject Headings

Image processing--Digital techniques; Algorithms; Computer graphics; Rendering (Computer graphics); Signal processing--Digital techniques

Publication Date

5-9-2003

Document Type

Thesis

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Shaaban, Muhammed

Advisor/Committee Member

Zeck, Norm

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TA1637 .S658 2003

Campus

RIT – Main Campus

Share

COinS