This book developed out of a series of publications in the area of image
processing with massively parallel algorithms. The topic of image
processing is a particularly promising area for the use of synchronous
massively parallel or data-parallel compu- ter systems which work
according to the SIMD principle (single instruction, multiple data).
While the era of large SIMD super-computers has passed, SIMD systems
have come back as dedicated vision subsystems and will soon be found
even in embedded systems. In comparison to conventional sequential
implementations of basic image opera- tions, this book illustrates the
intrinsic parallelism which is almost always present in image
processing. By utilising parallel algorithms it is even possible to
illustrate oper- ations in a simpler and easier to understand way than
for the sequential case. The presentation method chosen for this book
assumes that short, terse excerpts of program code will significantly
enhance the understanding of the material, e.g. of image operations,
while longer listings are more likely to distract from the topic. For
this reason, each chapter will not only define and explain the central
image processing algorithms with the help of examples, but will also
give an excerpt of a massively par- allel program. For image processing
this means that at least virtually there should be one processor
available for each pixel. The mapping onto a smaller number of exist- by
compiler, and as of ing real processors is done transparently the such
is not interest here.
