Implementation and use of software scanning on a small CLIP4 processor array
For a small CLIP4 processor array, algorithms are described that perform scanning over large images. Two algorithms are shown: a shift out algorithm developed by Stonefield Systems PLC., and a shift through algorithm, which is presented here for the first time. A simple hardware alternative to shifting, half scan addressing, is shown to offer a significant speed advantage.
It is shown that software scanning suffers from a severe drawback: it has a high overhead for simple neighbourhood operations. However, techniques are available to reduce the overhead to values, comparable with those of the hardware scanning CLIP4S system. These require the user to know the neighbourhood size needed for his operations, and the maximal neighbourhood allowed (a system constant). For point- and complex neighbourhood operations, software scanning can be faster than hardware scanning. An advantage of software scanning is its flexibility (in image size and sorts of operations), which also offers the possibility to do data dependent processing and recursion.
It is concluded that software scanning can be an alternative to hardware scanning, especially when flexibility is of value.
Unable to display preview. Download preview PDF.
- 1.Pass S.; The GRID computer system. In: Kittler J., Duff M.J.B. (eds.); Image processing system architecture, Research Studies Press, Letchworth, 1985.Google Scholar
- 2.Potter J.L.; MPP architecture and programming. In: Preston K. jr., Levialdi S., Duff M.J.B. (eds.); Multicomputers and image processing, Academic Press, London, 1982.Google Scholar
- 3.Duff M.J.B.; The CLIP4. In: Fu K.S., Ichikawa T., Special computer architectures for pattern processing, CRC Press, 1982.Google Scholar
- 4.Fountain T.W., Postranecky M., Shaw G.K.; The CLIP4S system, Pattern recognition letters 5, 1987.Google Scholar
- 5.Fountain T.W.; private communication, 1987.Google Scholar
- 6.Buurman J.,1987; Scanning algorithms for the CLIP4 processor array. Thesis, Pattern recognition group, Department of applied physics, Delft University of Technology, 1987.Google Scholar
- 7.Arcelli C., Cordella L., Levialdi S.; Parallel thinning of binary images, Electr. Lett. 11, 1975.Google Scholar
- 8.Hilditch C.J.; Comparison of thinning algorithms on a processor array. Image and vision computing 3, 1983.Google Scholar