Automatic character-reading device

Claim 1

Device for the automatic reading of characters filled in in boxes on a document, which device comprises a camera, as well as means for the digitalized transfer to a recording store of the character pattern information taken, a processor of the stored information, a features deriving device and classification means which determine the class of the characters with the aid of a probability table, characterized in that the character pattern information in the recording store (3) has a quantification of 64 x 64 image points per box area; in that the processor (4) of the stored information is provided with 1. an image separator (7) for scanning the information of the total area of all the boxes to record the image of each character pattern in an image area of 128 x 64 image points,the images being shifted against corresponding edges of the image areas, 2. a decision circuit (8), which, on the ground of the result of the image separator (7) (separation successful - separation unsuccessful), allows to discriminate between typewritten characters and handwritten characters, 3. a geometric separator (9) for separating the images of the handwritten characters, starting from four images each having 32 x 32 image points, which correspond with the image points from the original image area of 64 x 64 image points which are homogeneously distributed over the total image area, and which together form said image area, which four image areas are looked upon with such an edge that each area looked upon comprises 48 x 48 image points, of which four images thus extended the OR-function is determined for the geometrical separation, and 4. an intermediate store (10) for recording the information from the geometric separator (9) or from the image separator (7), if in at least one of them the separation has been successful, in which intermediate store the information can be found in four images of 32 x 32 image points against the left-hand and the upper edge, and also means are provided which determine on the ground of the dimensions of the character pattern if the image can be enlarged and, consequently, enhanced and, if this is the case, to execute the enhancement by making use of a preset combination of two or more images; in that the classification means (6) are such that on the ground of the information from the decision circuit (8) the class of the characters can be determined by means of a probability table for typewritten characters.and a probability table for handwritten characters.