Information can be recovered beyond direct visual readability in images
Beyond Readability: Evaluating Task Information Recoverability
Computer Vision and Pattern Recognition
Summary
Sometimes you can’t read all the details in a picture because it’s blurry or damaged. This paper shows that even when you can’t directly see the information, you might still be able to figure it out by recognizing what the picture is and looking up facts about it from a known source. The authors tested this idea by trying to identify book authors and titles from covers when some details were hard to read. They found that these extra clue-based methods can still work well even when direct reading fails, so it’s useful to measure both ways when checking how much information can be recovered.
What this means in practice
- •For mobile app developers: Create apps that extract information from damaged images by combining direct reading and knowledge-based recovery for better data extraction accuracy.
- •For document digitization teams: Improve recovery of bibliographic information from poor-quality scanned book covers by combining visual recognition with catalog lookups under degraded conditions.
Tested on one dataset.
Authors
Yiwei Liu
Abstract
Direct visual readability and task-information recoverability are different quantities. Failure to decode a target from a fixed observation need not eliminate access to that target through another recovery route. We develop an evaluation perspective that makes the observation, query, target, and available knowledge explicit and measures the overlap between routes' success sets. For information available on the original visible surface under suitable imaging conditions, direct optical recovery reads the target from the image, optionally after restoration; entity-linked recovery uses residual visual evidence to identify the depicted entity and accesses its target through an entity--attribute relation in a specified knowledge resource. Such access can draw on stored knowledge or an external source. A controlled book-cover study instantiates external access with a fixed title--author catalog, comparing optical author recovery with visual title resolution and deterministic lookup under resolution degradation. Entity-linked successes persist across the tested vision--language models, revealing information access beyond the tested direct visual frontier despite substantial differences in absolute performance. A substantial optical-only region remains. These complementary outcomes show why visual degradation should be evaluated through the task information accessible along specified routes and knowledge resources, alongside direct readability.