UC Berkeley CS294-9 Fall Document Image Analysis Lecture 20: Intro to Layout Richard J. Fateman Henry S. Baird University of California – Berkeley Xerox Palo Alto Research Center

UC Berkeley CS294-9 Fall Page layout analysis Structural (Physical, Geometric) Layout Analysis Functional (Syntactic, Logical) Layout Analysis Read-order determination

UC Berkeley CS294-9 Fall Structural Isolation of columns, paragraphs, lines words, tables, figures. Maybe letters. Without some layout analysis, much of the previous work would be impossible! Without layout analysis, what is the sequence of words in a multi-column format?

UC Berkeley CS294-9 Fall Functional Typically domain dependent May require merging or splitting of syntactic components Encoding into ODA (object oriented document architecture) or SGML (DTD describes components like section, title..)

UC Berkeley CS294-9 Fall Functional Components First page of a technical article may have Title Author Abstract, body/column1 body/column2 footnotes Pagination Journal name/volume/date… Business letter might have Sender Date Logo Recipient Body Signature

UC Berkeley CS294-9 Fall Finding structural blocks

UC Berkeley CS294-9 Fall Common Approaches Top Down analysis –Horizontal and vertical profiles –Recursive: columns, paragraphs/lines/words –As illustrated earlier Bottom Up analysis –Use adjacency based on Pixels / morphology of dilation (millions) RLE/ merge lines (thousands) Connected Components (hundreds) Look at the background (shape-directed covers) Also, human hints.

UC Berkeley CS294-9 Fall Standard images…the Scanned Input

UC Berkeley CS294-9 Fall Smear character boxes

UC Berkeley CS294-9 Fall Smear words to get lines

UC Berkeley CS294-9 Fall Smear lines to get paragraphs

UC Berkeley CS294-9 Fall Issues: Sensitivity to noise. Solutions: –Clean up via kfill or similar filtering, ruthlessly –Divide the page (artificially) and keep the noise from affecting the document globally Slanted lines. Solution(s): –Deskew (since it is not too hard(?)) –Use nearest neighbors “docstrum” Concave regions (text flow around a box). Solution(?) look at background Variation in font, spacing can throw off analysis –Allow for local analysis

UC Berkeley CS294-9 Fall Interactive semi-automatic zoning (RJF)

UC Berkeley CS294-9 Fall Zoom in

UC Berkeley CS294-9 Fall Scroll around

UC Berkeley CS294-9 Fall View individual pixels

UC Berkeley CS294-9 Fall Semi… Turn up the noise filter until we start to kill some of the punctuation. How? As we turn up the threshold, the number of connected components drops, then reaches a stable plateau after the noise is gone, and then drops again as we remove punctuation, the dots above the “i” etc.

UC Berkeley CS294-9 Fall auto… Turn the horizontal smear knob until the number of components drops suddenly from about 3000 to about 600. Character boxes have been merged into wordboxes Turn the horizontal smear knob until the number of components drop from about 600 to about 100. Wordboxes have become lineboxes.

UC Berkeley CS294-9 Fall matic.. Tweek the vertical smear knob. Lines become paragraphs. (Turn further, and paragraphs become columns).

UC Berkeley CS294-9 Fall Specify read order

UC Berkeley CS294-9 Fall Interactive functional tagging: mark subject/author/etc? Here we attempt automatic id of math… Automatic math zone. This is a challenge because the zone is in two parts, containing the math … f(p)=F(p)

UC Berkeley CS294-9 Fall Docstrum/ L.O’Gorman 5 nearest neighbors (ogorman93)

UC Berkeley CS294-9 Fall Example of “spectrum” Each point represents distance and angle of a cc. N^2, but not so bad.

UC Berkeley CS294-9 Fall Statistics for skew and spacing

UC Berkeley CS294-9 Fall Extract Lines, group to paragraphs Statistically close enough horizontally to be words, then lines Statistically close enough and parallel enough and the same length as… group two lines into the same text block. (arguably saving time by not deskewing; dealing with non-constant skew) Example follows..

UC Berkeley CS294-9 Fall Sections with different skew 6 business cards, nearest neighbors vectors

UC Berkeley CS294-9 Fall Extracted text lines, blocks Useful? General?

UC Berkeley CS294-9 Fall Without layout analysis Reading across columns Misplacing captions Misplacing footnotes Misunderstanding page numbers (which should be REMOVED in the reformatting process) Need extraction of biblio data: title, author, abstract, keywords Nearly every subsequent step is compromised by lack of context.

UC Berkeley CS294-9 Fall A Diversion: Separating Math from Text Why separate math from text? Types of mathematics encountered Previous Work Two approaches –post-processing commercial OCR –character-based (details!) Errors and their correction Ambiguities

UC Berkeley CS294-9 Fall Why separate math/text/images/.. OCR programs do not work for math becomes, in Textbridge, Designation as a “picture” is only a partial solution

UC Berkeley CS294-9 Fall Mathematics on a Page Inline is harder to pick out because it may look like italics text

UC Berkeley CS294-9 Fall Previous Work Isolation by hand (most math parser papers) Texture/ statistics based heuristics –useful for display math “paragraphs” –not useful for in-line math Character based pseudo-parsing (but without font information or true parsing feedback) Incomplete

UC Berkeley CS294-9 Fall Proposal: Post-Processing of OCR Start with commercial best-effort recognition Reprocess the intermediate data structure (e.g. for TextBridge, the XDOC file) Accept recognition of text zones with high recognition certainty. (Lines with no errors surrounded by lines with no errors are considered solved)

UC Berkeley CS294-9 Fall Separate uncertain areas Re-consider “the rest of the image” as potential mathematics zones: uncertain regions (including nearby “certain” characters/lines) Isolate characters, identify fonts, etc. Play out heuristic rules for separating text and math zones. Consider eradicating math and re-submitting text; separately recognizing math and reinserting in XDOC

UC Berkeley CS294-9 Fall Alternatively, Starting from our own naïve OCR Connected component recognition Separate characters by initial classification Repeatedly re-examine via rules Determine text zones, remove math / feed remainder to commercial OCR –How best to blank-out math? XXX Most likely human interaction remains

UC Berkeley CS294-9 Fall Two bags: Math vs Text Initially MathInitially Math –+ - = / Greek, scientific symbols, 0-9, italics, bold, (), [], sin, cos, tan, dots, commas, decimal points Initially TextInitially Text –Roman Letters, junk

UC Berkeley CS294-9 Fall Sample Text Bag

UC Berkeley CS294-9 Fall Sample Math Bag

UC Berkeley CS294-9 Fall Second Pass Correct for too much Math Grow “clumps” (expand BBs) to categorize – vs “end of sentence.” –(comment) vs f(x) –hyphen-words vs x 2 - y 2 –horizontal lines generally –isolated 1 or is it l “ell” or I “eye” “bags” or “zones” of geometric-relation boxes containing either words or potential math

UC Berkeley CS294-9 Fall Importance of Context Here are 12 L’s and a 1

UC Berkeley CS294-9 Fall Third Pass Too much is in the text bag now –blur the math to allow for embedded Roman text like “sin” or “l” Re-clump the mathematics to see if new bridges have been formed Some italics in the math bag may be really –English words in theorems –emphasized text

UC Berkeley CS294-9 Fall On Ambiguity and Correctness Can we find the math in ad - bc by ad hoc methods? If we are unable to disambiguate English words, why should we be able to disambiguate mathematics? Abuse of mathematical notation is widespread: can we insist that new papers either have a non-ambiguous notation or an underlying electronic non-ambiguous notation?

UC Berkeley CS294-9 Fall Conclusions We can make a first cut on separating math from text If we wish to “enliven” math publication with semantic underpinnings, this may help in their production Incorporation of AI rule-based transformations as well as hand correction are likely to be important