Manual Correction of Misclassified Image Segments

Since book scanning is often used for low-volume titles, controlling scan cost is critical. Automatic image segmentation promises high quality and low cost scans with a minimal need for user intervention. However, since no segmentation algorithm is perfect, strategies based on business policies are required to manage trade-offs between cost and quality. We describe an adaptive approach that allows the user to control, based on policy, the amount of intervention necessary to verify the segmentation's correctness and to rapidly adjust misclassified segments.