The Effects of Bandwidth Compression on Image Interpreter Performance
Harold Martinek, Alvin Zarin · 1979
Conventional imagery differing in resolution (8-inch ground resolved distance GRD, 16-inch GRD, and 24-inch GRD), digitized and treated to represent four levels of bandwidth compression (1:1=no compression, 4:1, 8:1, and 10:1) was interpreted by 12 experienced image interpreters. A Greco-Latin square design was used to control for differences between interpreters, imagery subsets, and period effects at each GRD and to test the effects of bandwidth compression. The dichotomized image variables of sun angle, image contrast, and target obscurity were distributed equally across all conditions, but could be analyzed only for the 24-inch GRD imagery. Significant decreases were found in the number of correct identifications due to bandwidth compression; the largest decrease occurred between 4:1 and 8:1 bandwidth compression ratios. System developers should make careful trade-off evaluations in using bandwidth compression. More research in this area is required on the search function of interpretation and interactions of image variables.