A Comparison of Whitespace Normalization Methods in a Text Art Extraction Method with Run Length Encoding

Abstract

Text based pictures called text art or ASCII art can be noise in text processing and display of text, though they enrich expression in Web pages, email text and so on. With text art extraction methods, which detect text art areas in a given text data, we can ignore text arts in a given text data or replace them with other strings. We proposed a text art extraction method with Run Length Encoding in our previous work. We, however, have not considered how to deal with whitespaces in text arts. In this paper, we propose three whitespace normalization methods in our text art extraction method, and compare them by an experiment. According to the results of the experiment, the best method in the three is a method which replaces each wide width whitespace with two narrow width whitespaces. It improves the average of F-measure of the precision and the recall by about 4%.