Rocr Recognition Point maps
To create a Recognition Point (RP) "map" for a character, you must specify a few salient points by which it can be uniquely identified.
When the sample image is displayed in the Refound – image document recovery dialog, it is automatically split into separate blocks wherever vertical lines of white pixels are found. Initially, each block is simply allocated a glyph number. The User sets up the appropriate character code when building the RP map; this 'Character' is shown in black.
To create a Recognition Point, simply click any black pixel on the edge of the character's glyph: the pixel turns red. The database stores the position of the selected pixel and the pattern of black and white pixels surrounding it (at least 1 of these 8 pixels must be white).
Only 4 or 5 RPs are usually needed to match glyphs in an imaged document to an individual "character". If there are several glyphs with similar contours, however, their maps may require more RPs.
To minimize the number of RPs in a map, consider choosing pixels where the stroke starts, ends, turns sharply or joins with another, and on curves. Try to avoid choosing pixels at the far left or right of the contour, as adjacent characters may merge at these points and the edge pattern lost.