Method and system for document image layout deconstruction and redisplay system

Abstract

The invention converts a document originating in a page-image format into a form suitable for an arbitrarily sized display, by reformatting or "re-flowing" of the document to fit an arbitrarily sized display device.

Description

[0001] 1. Field of Invention[0002] The invention relates generally to the problem of making an arbitrary document, conveniently readable on an arbitrarily sized display.[0003] 2. Description of Related Art[0004] Existing systems for rendering page-image versions of documents on display screens have required manual activities to improve the rendering, or clumsy panning mechanisms to view direct display of page images on wrong-sized surfaces. In particular, it has been necessary to either (1) key in the entire text manually, or (2) process the page images through an optical character recognition (OCR) system and then manually tag the resulting text in order to preserve visually important layout features.[0005] Problems with existing systems include: (a) high expense of manual keying and / or correcting of OCR results and manual tagging; (b) the risk of highly visible and disturbing errors in the text resulting from OCR mistakes; and (c) the loss of meaningful or aesthetically pleasing t...

AI Technical Summary

Benefits of technology

[0008] Reflowing is a process that moves text elements (often words) from one text-line to another so that each line of text can be contained within given margins. Reflowing typically breaks or fills lines of text with words, and may re-justify column margins, so that the full width of a display is used and no manual `panning` across the text is needed. As an example, as a display area, within which lines of text appear, is altered so that the width of the visible text is reduced, it may be necessary for words to be moved from one text-line to another to shorten the length of all of the text-lines so that no text-line is too long to be entirely visible in the display area. Conversely, if the display area is widened, words may be moved from one text-line to another so that the length of text-lines increase, thereby allowing more text-lines to be seen without any word image being obscured.

[0015] This approach to text line modeling has several advantages over known projection or linking methods. First, different text lines can have different orientations. Second, by taking into account both the baseline and the descender line, the technique can find text lines that are missed by known text line finders. Third, the matches returned by this method follow the individual text lines more accurately than other known methods.

[0016] Column boundaries are identified in a similar manner by finding globally optimal maximum likelihood matches of the center of the left side of bounding boxes against a line model. In order to reduce background noise, prior to applying the line finder to column finding, statistics about the distribution of horizontal distances between bounding boxes are used to estimate the intercharacter and inter-words spacing, i.e., the two largest components in the statistical distribution of horizontal bounding box distances. The bounding boxes for characters are then merged into words. This reduces severalfold the number of bounding boxes that need to be considered for column matching and tends to improve the reliability of column boundary detection.

[0018] This simple layout analysis technique copes with a large number of commonly occurring layouts in printed documents and transform such layouts into a sequence of images that can be reflowed and displayed on a smaller-area display device. The simple technique works well in these applications because the requirements of reflowing for a smaller-area display device, such as a document reader, are less stringent than for other layout analysis tasks, like rendering into a word processor. Since the output of the layout analysis will only be used for reflowing and not for editing, no semantic labels need to be attached to text blocks. Because the documents are reflowed on a smaller area screen, there is also no user expectation that a rendering of the output of the layout analysis precisely match the layout of the input document. Furthermore, if page elements, like headers, footers, and / or page numbers, are incorporated into the output of the layout analysis, users can easily skip such page elements during reading. Such page elements may also serve as convenient navigational signposts on the smaller-area display device.

[0025] Thus, redisplaying the document is enabled for a wide range of displays whose size, resolution, contrast, available colors, and / or geometries may require the document's contents to be reformatted, reflowed, re-colored, and / or reorganized to achieve a high degree of legibility and a complete understanding of the document's contents, without requiring OCR or re-keying, and without being subject to the respective attendant errors of OCR or re-keying, and without losing the look and feel of the original document as chosen by the author and publisher.

[0026] In various exemplary embodiments, the methods and systems according to this invention reduce costs by obviating the need for manual keying, correction of OCR results, and / or tagging. In various exemplary embodiments, the methods and systems according to this invention tend to avoid introducing OCR character recognition errors. In various exemplary embodiments, the methods and systems according to this invention tend to preserve typeface and type size choices made by the original author and publisher, which may be helpful, or even essential, in assisting the reader in understanding the author's intent. In various exemplary embodiments, the methods and systems according to this invention also tend to preserve the association of graphics and non-textual elements with related text.

Problems solved by technology

Existing systems for rendering page-image versions of documents on display screens have required manual activities to improve the rendering, or clumsy panning mechanisms to view direct display of page images on wrong-sized surfaces.

Problems with existing systems include: (a) high expense of manual keying and / or correcting of OCR results and manual tagging; (b) the risk of highly visible and disturbing errors in the text resulting from OCR mistakes; and (c) the loss of meaningful or aesthetically pleasing typeface and type size choices, graphics and other non-text elements; and (d) loss of proper placement of elements on the page.

Such problems are significant, for example, because book publishers are increasingly creating page-image versions of books currently being published, as well as books from their backlists.

While print-on-demand images can be re-targeted to slightly larger or slightly smaller formats by scaling the images, they cannot currently be re-used for most electronic book purposes without either re-keying the book into XML format, or scanning the page images using OCR and manually correcting the re-keyed and scanned images.

Images