Title :
Two-dimensional convolution on a pyramid computer
Author :
Chang, J.H. ; Ibarra, O.H. ; Sohn, S.M.
Author_Institution :
Dept. of Comput. Sci., Minnesota Univ., Minneapolis, MN
fDate :
7/1/1988 12:00:00 AM
Abstract :
An algorithm for convolving a k×k window of weighting coefficients with an n×n image matrix on a pyramid computer of O(n2) processors in time O(logn+k2), excluding the time to load the image matrix, is presented. If k=Ω (√log n), which is typical in practice, the algorithm has a processor-time product O(n 2 k2) which is optimal with respect to the usual sequential algorithm. A feature of the algorithm is that the mechanism for controlling the transmission and distribution of data in each processor is finite state, independent of the values of n and k. Thus, for convolving two {0, 1}-valued matrices using Boolean operations rather than the typical sum and product operations, the processors of the pyramid computer are finite-state
Keywords :
Boolean functions; computational complexity; computerised picture processing; matrix algebra; parallel architectures; 2D convolution; Boolean operations; computational complexity; computerised picture processing; finite state; image matrix; pyramid computer; window; Computer architecture; Computer vision; Concurrent computing; Conferences; Convolution; Labeling; Layout; Polynomials; Signal processing; Testing;
Journal_Title :
Pattern Analysis and Machine Intelligence, IEEE Transactions on