- 25 - 
4. Digital Processors for Image Processing 
Digital image processing done on a time sequential one-processor- 
computer of John von Neumann type is very time consuming. There- 
fore much effort has been devoted to study processor architec- 
tures for this type of operations /29/. Figure 31 shows several 
combinations of instruction and data streams through a processor 
system. Pigure 31a represents a single processor system where one 
processor is handling a single data stream. A system according 
to figure 31b has an array of processors and each processor is 
applying the same instructions to different data streams. Contrary 
to this mode of operation a single data stream is flowing through 
several processors where each processor is using its own program 
(figure 31c). Figure 31d shows a general multiprocessor system 
with parallel access of a number of processors to different 
partitions of memory. 
An analysis of typical image processing procedures like local ope- 
rations shows, that many operations could be done in parallel. This 
led to the development of array processors like ILLIAC IV and 
STARAN. Because of the large hardware expense of such processor 
structures the complexity of operations which can be done in paral- 
lel is very limited. Beyond this the input-output interface bet- 
ween the array processor and a commercial Storage medium covers 
a lot of problems. Especially in the case of local operations like 
convolutions a structure shown in figure 32 seems to be better 
Suited. In this case image data are flowing on a byte sequential 
bus from a commercial storage medium to a tapped data pipeline. 
From there the data are fed to different processor levels. Such 
a pipeline of array-processors allows to handle rather complex 
image operations within one Step. In order to handle a large varie- 
ty of problems it is necessary to make the structure of the pipe- 
line programable. An example of such a structure /30/ is given 
in figures 33 and 34 for the differentiation procedure being de- 
Scribed in chapter 3.1.2 (figure 15, equ. 12). 
5. Literature 
717 
Shulman, A.R. 
Cutrona, L.J. 
Vander Lugt, 
Rotz, F.B. 
Klooster, A. 
Leith, E.N. 
Kozma, A. 
Upatnieks, J. 
Feinleib, J. 
Oliver, D.S. 
Nisenson, P. 
Iwasa, S. 
Oliver, D.S. 
Buchan, W.R. 
Lipson, S.G. 
Nisenson, P. 
"Optical Data Processing" 
John Wiley and Sons, Inc. 
New York, London, Sidney, Toronto 
1970 
"Recent Developments in Coherent 
Optical Technology" 
Optical and Electro-Optical 
Information Processing; MIT-Press 
1965, S. 83-124 
"Character Reading by Optical 
Spatial Filtering" 
Optical and Electro-Optical 
Information Processing, MIT-Press 
1965, S. 125-142 
"Coherent Optical Systems for Data 
Processing, Spatial Filtering and 
Wavefront Reconstruction" 
Optical and Electro-Optical 
Information Processing, MIT-Press 
1965, S. 143-158 
Holographie 
Berliner Union Verlag, 
Stuttgart, 1973 
"Reusable Optical Image Storage 
and Processing Device" 
Applied Optics, Vol. 11, No. 12 
Dec. 1972, S. 27952-2759 
"Real Time Optical Processing 
with Bi, Si0,9 PROM" 
Applied épriés, Vol. 11, No. 12 
Dec. 1972, S. 2760-2707 
"An Optical Storage and Processing 
Device Using Electrooptic Zns" 
IEEE Trans. on Electron Devices, 
May 1974, S. 769-773 
"Imaging Characteristics of the 
Itec PROM" 
Applied Optics, Vol. 13, No. 9 
Sept. 1974, S. 2052-2060