Towards Versatile Document Analysis Systems

Abstract

The research goal of highly versatile document analysis systems, capable of performing useful functions on the great majority of document images, seems to be receding, even in the face of decades of research. One family of nearly universally applicable capabilities includes document image content extraction tools able to locate regions containing handwriting, machine-print text, graphics, line-art, logos, photographs, noise, etc. To solve this problem in its full generality requires coping with a vast diversity of document and image types. The severity of the methodological problems is suggested by the lack of agreement within the R&D community on even what is meant by a representative set of samples in this context. Even when this is agreed, it is often not clear how sufficiently large sets for training and testing can be collected and ground truthed. Perhaps this can be alleviated by discovering a principled way to amplify sample sets using synthetic variations. We will then need classification methodologies capable of learning automatically from these huge sample sets in spite of their poorly parameterized-or unparameterizable-distributions. Perhaps fast expected-time approximate k-nearest neighbors classifiers are a good solution, even if they tend to require enormous data structures: hashed k-d trees seem promising. We discuss these issues and report recent progress towards their resolution.

Feature selection focused within error clusters

“…We conduct experiments in the document image content extraction framework [10,1]. In this framework, each pixel is treated as a sample.…”

“…We conducted the experiments in a document image segmentation framework, by trying to classify pixels into machine-print or handwriting [10,1]. A sequence of six classifiers trained with the augmented feature sets exhibits a monotonically decreasing error rate.…”

Issues in Performance Evaluation: A Case Study of Math Recognition

“…For example, Baird and Casey report that the three CD-ROM University of Washington database cost about two million dollars [7][20]. Copyright considerations can be a barrier to making a data set available e.g.…”

Document Content Extraction Using Automatically Discovered Features

“…We have applied our algorithm to a document image content extraction problem: finding regions containing machine-printed text, handwriting, photographs, etc in images of documents [4,5,2,1]. We select a small window around each pixel as our sample space and look for some linear combination of pixel values within this window as our new feature.…”