Document image understanding
Sargur N. Srihari · 1986
/f rEihl document image is an optically scanned and dryesentation of aprintedpage thatconsisb of blocls hdrawings, half-tone pictures, and icons. Document dastanding is a goal-oriented problem involving f d interpreting differsnt blocks and coordinating the to achieve an end result. We examine several rc nbproblems in document understanding tasks. The lr rmge from pixel processing issues to symbolic xning to global control tttcnt image is a visual representation of a printed o a journal article page, a magazine cover, a IEge, etc. Typically, it consists of blocks of text, tadsr and sentencei that are intenpersed with half line drawings, and symbolic icons. A digital rltP p is rs a two-dimensional array representation ofaa 4cobained by optically scanning and raster Lrd zd coov copy do documenl It may also be an electronic,rrs crcated in that form, say, for a bit-mapped [rcr printer. An example of a document image is Ercl. i tge understanding (or simply, document b r a goal orienled problem that may involve Ihotographs l5oographs otocraDhs (which (whrch is ls a visiaz vrsron problem), problem). (which. may involv.e natur-al language interpreting line drawings such as graphs, in fuinteractions b interactions of the different components is Sore examples of document understanding m hEgrate the pictorial content of a newspaper rflamoanving eqqpanymg ing narrative narrahve'title. title, (ii) determine the tin n address on a colorfirl magazine cover: it is adqientthe mostplausible textual block, o-'-t it has the correct syntax, and (iii) make motated.x-y plot. dcrsanding implies the correctrccognition of Et Ent comDonents- components. The rccosnition recognition has to be fth level components such as columns and Ias in teirrs of low level components such as Ite case of text; curyes, rectangles, and solids and image regions in case of half-tone clrsis is the task of deriving a high level cmEnts of a document image. The spatial can be represented in several ways. n of the image of Figure 1 are shown in ft- a analogical representatio-n-that gives the including paragraphs and the second is a ation in the form of a partial semantic represent document entities and arcs G qL Each of them assumes blocks as units lbe associated with deails like size, shape, lirels and texture information. This-- d]----,F