User Strategies in Interactive Image Processing

Abstract

Several related experiments are being conducted to investigate human performance in an image processing and interpretation task. In this task, users select operators from a set of image enhancement transformations in order to modify an image and facilitate interpretation. In several trials, subjects were required to identify images of ship hulls as belonging to one of four categories. Operator sequences, selection times, and categorization responses were recorded. Subjects adopted strategies that lead to the enhancement of spatial frequencies important for the categorization task. The strategies were easily identified, emerged in an orderly manner, and changed in predictable ways in response to changes in the frequency content of the task images. We are currently developing a production-based model to account for the operator sequences and selection times.