l?i3dSMi,P.~,M.ROtti UniuemlW deali Sbdi di Mlano - Dipar!i~ Vit%vbtti5 20133 Milano, KALY di FIska many scientific and technical environments observationaldata areoftena&c&l in the form of digital images.For example,physiciansobservetheir patientsvia CT devices [Ca87], while astronomers exploit CCD devices [Ac89b] and environmentor agricultural experts exploit Landsatimagesra87]. Inalltheseca3e3dataarecollectedandorganizedasdigital images,which are finite bidimensional arrays of integer numbers.It is important to note that looking at an array as a digital imagemeansto look at its elementsas pidure elements(or pixels) whose meaning is determined not only by their numeric values,but even by their (mutual) positions, ie. their topological and geometrical properties. It has loq been recognizedthat APL is the obvious tool for the dcfmition 8nd description of the algorithms computing these propertiesand used in such activities as ima& procazssing fBrsO], analysisof pictorial data jBi81) or vision experimentspe88]. In this paper we argue, on the basii of our experience [Ac89a] [CUSS][De84], that APL notation (in our casean adaptation of the aw notation of AFU) allows the combination of tools to defme image interpretation strategies which detectsignificant entities appearingin an hage and classifythem as the tracks of objectsin the real world. In a digital imagethe sigu%ant entities appear as setsof almilaf pixels (numbers), called structures.In order to
/lp/association-for-computing-machinery/definition-of-image-interpretation-strategies-in-apl-aPHZaTZRaY