Printer-friendly page - just print

Using the Ntuplib Library (Introduction Excerpt)

Andrew Shapira

ECSE Department
Rensselaer Polytechnic Institute
Troy, NY 12180


Note – this is the introduction and the one-entry bibliography from the document Using the Ntuplib Library that is distributed with the Ntuplib library.


This document describes how to use the Ntuplib library. The Ntuplib library is a C library for finding n-tuples. An n-tuple is a set of n pixels that is designed to distinguish two character classes. One class is termed termed the positive class and the other the negative class; together, the classes constitute a dichotomy. The Ntuplib library finds n-tuples that fit characters in the positive class, and do not fit characters in the negative class. The library forms a foundation for pattern recognition experiments like the ones reported in [1]. For further details, including how the tuples found by Ntuplib are resistant to noise and sampling variations, see [1].

An outline of the remainder of this document is as follows. Section 2 discusses how to use the nt utility. The nt utility is an executable program that serves as a front end to the library. Users can generate distinguishing tuples by invoking nt from the operating system; in UNIX a typical invocation would come from a shell. Section 3 discusses a few miscellaneous items. Installing the library is discussed in Section 4. A sample run of the nt utility may be found in the appendix.

There is currently no documentation that describes how to invoke Ntuplib from programs other than nt. One way to figure out how to do this is to look at the nt source code, in conjunction with this document. Possibly, a future version of this document will contain more information about this, although this is not at all certain.

To understand this document, the reader should be familiar with the main concepts of [1]. Some details of the tuple generators may be found there.

In this document, as well as in the source code, the back1 and back2 generators refer to the generators that are respectively called Gen0 and Gen1 in [1]. The back1 generator exists mainly to serve as a benchmark for the more powerful back2 generator.


[1] D. Jung, M. Krishnamoorthy, G. Nagy, A. Shapira, N-Tuple Features for OCR revisited, IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995.