Connected Component Level Discrimination of Handwritten and Machine-Printed Text Using Eigenfaces

Abstract

We employ Eigenfaces to discriminate between handwritten and machine-printed text at the connected component (CC) level. Normalized images of machine print CCs are treated as points in a high-dimensional space. PCA yields a reduceddimensional character space. Representative machine print CCs are projected into character space and a local distance threshold for each representative is automatically determined. CCs are classified as machine print if they are within the local distance threshold of their closest machine print representative. Otherwise, they are classified as handwriting. Recursive character segmentation using min graph cut is used to address the problem of touching characters. Validation over a large NIST handwriting and machine print database demonstrates precision of 93.98% and 89.1% for machine print and handwriting respectively.

Handwritten and Machine-Printed Text Discrimination Using a Template Matching Approach

“…The variations of machine-printed samples are significantly lower than those of the handwritten class. As a result, the feature space extracted from the machine-printed samples are more concentrated, while the same features of the handwritten text samples are mapped to a significantly wider range [2]. This fact has been employed by many of the previous approaches to assign classification boundaries between the two classes, which provided acceptable results over particular types of documents.…”

“…An Eigenface-based approach has been proposed for the HMC task in [2]. As in [15], first, principal component analysis (PCA) is used to create a set of normalised characters in different font styles.…”

Pixel Level Handwritten and Printed Content Discrimination in Scanned Documents

“…The classification of single characters is also possible. An interesting approach is proposed by Pinson et al [8], i.e., the use of "eigenfaces" adapted to text, which lead to a mean recall of 91.5%. Koyama et al [9] presented a method based on frequency domain analysis can reach a precision of 97% on different kinds of characters.…”

Multiscale Region Projection Method to Discriminate Between Printed and Handwritten Text on Registration Forms

“…Handwriting may need to be extracted from forms and processed independently whereas machine-print-connected components (CC) are passed to an optical-character-recognition engine. 28 Registration forms are special documents that can be viewed as consisting of layers (letterhead, content, signatures, handwritings, table line, stamper, noise, etc.). Document analysis segments a registration-form document into layers with di®erent physical and semantic properties.…”