Identification of Very Similar Filled-in Forms with a Reject Option

Abstract

In this work, a technique addressed to the reliable identification of very similar filled-in forms, with a reject option, is proposed. The method is based on the automatic detection of the most discriminant regions at the image level, to be used by a distance-based classifier. Experiments included multi-page form images and the results suggest that a very high accuracy can be achieved when identifying previously known types of forms, even when unpredictable fillin data and significant noise from the scanning process appear on the test images. Likewise, it is shown that the probability that an unknown type of document is not rejected is very low. The presented technique could also be applied to other kinds of documents.

Handbook of Document Image Processing and Recognition

“…Thus, each document class can be associated to a discriminating landmark area, as a sub-image that can be used to discriminate the class from the others. The interest of these features is that it can be used for both identification and reject of forms [16].…”

“…Data Organization Once features have been extracted, it is necessary to organize them for the classification. The usual methods can be used, such as similarity measures between histograms [18], and feature vectors or matrixes [12,13,16,17,19].…”

Two Methods for Filled-in Document Image Identification Using Local Features

“…A method to select a set of potentially discriminant reference points to be used in a document identification task is presented in section 3.1. This method has been used to extract features and classify documents by means of two approaches: a new incremental version of the method proposed in (Arlandis et al, 2009), based on the cross matching between pairs of documents (section 3.2), and the method proposed in (Arlandis et al, 2011), which relies on the combination of the evidence contributed by multiple local features and a direct voting scheme (section 3.3).…”

“…This is an incremental version of the method presented in (Arlandis et al, 2009). It reduces drastically the time needed to train a high number of document classes of that method, particularly when the documents to be indexed are not very similar.…”

“…On the other hand, in the presented method, a local feature is selected based on its discriminative power between pairs of images, instead of requiring to find it discriminitive among the rest of the classes. Similarly to (Arlandis et al, 2009), this approach relies on the fact that the discriminant power of a subimage of a fixed size I c…”