The tremendous variety of image processing systems and algorithms cries out for some standardized way of comparing them--a benchmark. The author of this paper was charged with the task of devising the benchmark, which became known as the Abingdon Cross, after the town in Great Britain where the Abingdon workshop on architectures, algorithms, and languages for image processing was held in 1982.
The Abingdon Cross benchmark test image consists of two bright strips crossing at right angles on a dark background [1]. In the common area where the strips cross, the brightness is the sum of the brightnesses of the two strips. A white Gaussian noise with zero mean is added at all points of the image. Unlike certain images frequently used to compare or illustrate the results of one or another algorithm, such as the well-known baboon or the girl in a hat, the definition of this benchmark allows anyone to reproduce it without the necessity of procuring a tape or disk. The reference image has dimension 512 × 512 × 8.
This note gives the results of testing 51 machines in nine countries. Ratios of price to performance had a 1000:1 range, illustrating “the enormous increase in performance per dollar over the past decade without a sacrifice in speed.” This paper includes a broader range of tested systems than Preston’s earlier publication [1].