Abstract

Image filtering with the larger, and potentially most valuable, digital filters is very time-consuming, thus precluding use of these filters in routine clinical applications. A recently developed algorithm for spatial-domain filtering is described, and its speed is compared with those of conventional methods with and without an array processor. Using the new Chebyshev method, a 64 by 64 pixel image can be filtered on a standard 16-bit minicomputer with filters of size 3 by 3 to 23 by 23 in 1.4-9.2 sec. The conventional spatial-domain algorithm requires 3.8-71 sec. With an array processor, filtering is accomplished in 0.19-0.54 sec. Filtering in the frequency domain requires 34 sec without an array processor and 0.12 sec with one. Thus with this new Chebyshev algorithm, clinically practical digital filtering can be performed with large filters even without an array processor.