Browse Prior Art Database

Improving Readability of Cursive Characters Through Image Recognition

IP.com Disclosure Number: IPCOM000243983D
Publication Date: 2015-Nov-03
Document File: 3 page(s) / 47K

Publishing Venue

The IP.com Prior Art Database

Abstract

Described is a method in which rendered electronic documents that contain cursive text can be post-processed to clear common ambiguities.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 47% of the total text.

Page 01 of 3

Improving Readability of Cursive Characters Through Image Recognition

Languages that are written only in cursive can be difficult to display on electronic publications due to the flowing nature of the words. This is because the typefaces need to be designed in such a way to allow for some characters to connect to any of a number of possible neighboring characters. Because cursive characters often differ depending on whether the character is alone, or at the beginning, middle, or end of a word, producing a single connection point that correctly links with any neighboring character is, at best, challenging and, at worst, not possible. The result is that some of these connections can appear unnatural to readers, inhibiting the readability of documents in cursive-only languages.

    When a document is rendered in a reader (web browser, PDF viewer, e-book reader, etc.), the rendered document is stored as an image in a frame buffer. This invention adds the ability for reader software to provide the video driver with the extents of the document that it has typeset and rendered for display. Image recognition is then performed in the video driver (in the identified regions only) to identify typeface flaws that impact readability. The rendered document is modified in the frame buffer on the fly to correct the flawed characters, without performing any modifications of the underlying representation of the document.

    This approach is designed to make subtle corrections to individual characters in rendered cursive to make the document more readable; it is not suitable for making large-scale changes that may affect the length of words, spacing between words, etc., all of which could negatively impact the typesetting of the document. The utility of this idea is best served with a few examples.

Example 1:

    This image is a link from a Wikipedia* article, indicating that the article is available in Persian:

    The Persian word (read right to left) is "Farsi". There is a problem with the connection between the last two characters, "seen" (the character that looks like a "w") and "yeh" (everything to the left of the "w"). The upward hump should not exist; the upward motion implies that there is an additional, indistinct character between the seen and yeh characters, or that the arabesque in the yeh character is actually a different character. The difficulty is compounded when the font is small, as the reader may be concerned that he or she is missing some small dots or other diacritical marks (the above example is enlarged 150%, making it clear that there are no such marks). In this typeface, the yeh looks much like a "beh" character ( ) that is missing a dot; a reader is likely to stumble, squinting to look for the dot, or even mistaking a speck of debris on the screen for the dot and ending up with a nonsense word. Even if the font is large enough to prevent confusion with another character, the word would cause a Persian reader to pause, expending extra ti...