As mentioned previously, the algorithms presented in Sects. 2.1 and 2.2 are suitable for pipeline processing. They may be implemented in either commercially available short-pipelines for image processing (e.g., those marketed by Gould, Vicom, Grinnel, etc.), or completely new pipeline architectures composed of both general- and special-purpose hardware components. Pipeline architectures are well known for their ease of interconnection [Kent85]. In this chapter, we concentrate on a new pipeline implementation for Radon-based image processing. However, other image-oriented architectures have been explored. Particularly appealing are the MIMD/SIMD approaches considered in [Rice85] and [Silb85], M/SIMD machines [Ibra85], and systolic networks [Hara85], among many others.
Unable to display preview. Download preview PDF.