THE EFFECTS OF DOCUMENT IMAGE DEFECTS ON LINE DRAWING ANALYSIS ALGORITHMS
Yuh-Lin Chang, Daniel Lopresti · 1994
Image defects and their effects on drawing analysis algorithms are investigated in this work. To study general drawing analysis systems, we use unconstrained, well-behaved random polygons as test inputs. We generate synthetic noisy samples through the use of image defect models. Image analysis algorithms are then applied to these samples, and the results are empirically evaluated by matching them against a ground truth. Discrepancies are regarded as image analysis errors and categorized. The relationship between image defects and an algorithm's error behavior can then be analyzed. Our study has applications in the development of robust and reliable document image understanding systems.