Files in this item
|(no description provided)|
|Title:||Image Processing on Mpp-Like Arrays (Massively Parallel Processor)|
|Author(s):||Coletti, Neil Boyd|
|Department / Program:||Computer Science|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Abstract:||The desirability and suitability of using very large arrays of processors such as the Massively Parallel Processor (MPP) for processing remotely sensed images is investigated. The dissertation can be broken into two areas.
The first area is the mathematical analysis of emulating the Bitonic Sorting Network on an array of processors. This sort is useful in histogramming images that have a very large number of pixel values (or gray levels). The optimal number of routing steps required to emulate a N = 2('k) x 2('k) element network on a 2('n) x 2('n) array (k (LESSTHEQ) n (LESSTHEQ) 7), provided each processor contains one element before and after every merge sequence, is proved to be 14 SQRT.(N) - 4log(,2)N - 14. Several already existing emulations achieve this lower bound.
The number of elements sorted dictates a particular sorting network, and hence the number of routing steps. It is established that the cardinality N = 3/4(.)2('2n) elements used the absolute minimum routing steps, 8 SQRT.(3) SQRT.(N) - 4log(,2)N - (20 - 4log(,2)3). An algorithm achieving this bound is presented.
The second area covers the implementations of the image processing tasks. In particular the histogramming of large numbers of gray-levels, geometric distortion determination and its efficient correction, fast Fourier transforms, and statistical clustering are investigated.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1983.
|Date Available in IDEALS:||2014-12-15|