Abstract: When an image is entered from an input means, a line extracting means extracts and smoothes the contour of that image to obtain a tangent line passing through a certain point on that contour. A region with the certain point as the origin is determined, where Hough transform of a straight line is carried out with the center point of that region as the origin. Hough transform is carried out for all the points adjacent to the origin to obtain an envelope. A tangent line is predicted by the obtained envelope.

Abstract: In a character recognition system, a method and apparatus for segmenting a document image into areas containing text and non-text. Document segmentation in the present invention is comprised generally of the steps of: providing a bit-mapped representation of the document image, extracting run lengths for each scanline from the bit-mapped representation of the document image; constructing rectangles from the run lengths; initially classifying each of the rectangles as either text or non-text; correcting for the skew in the rectangles; merging associated text into one or more text blocks; and logically ordering the text blocks.

Abstract: A system for representing line drawings by polygonal lines. In the system, a binary digital image is scanned in a main scanning direction to form picture element connection data with a first image picture element as a start point picture element which is scanned next to a first background picture element and with a second background picture element as an end point picture element which is scanned next to a second image picture element, and the relationships between the picture element connection data are detected to provide as picture element block data a group of picture element connection data which have predetermined relationships.

Abstract: A control computer, an error calculator, a Bezier curve divider and an output device are provided for generating a character curve. Coordinates of end points and control points of a Bezier curve are input to the control computer. The error calculator determines squares of distances of the input control points from a straight line connecting the end points as errors. If these errors exceed a tolerable range, the curve divider divides the curve into two parts, and new coordinates of new end points and control points of a divided Bezier curve are input to the error calculator. The curve division and error calculation are repeatedly carried out until the calculated errors fall within the tolerable range. The end points of the respective divided curves are connected by straight lines to develop the original curve.

Abstract: An apparatus and method useful in correcting character strings misread by a character recognition engine. The misread string is compared to character strings contained in a lexicon of valid character strings and the similarity between the strings measured. The valid strings can be ranked by degree of similarity should further selection be required to determine the appropriate value of the misread character string.

Abstract: A matrix calculating circuit for calculating with respect to a matrix in which all diagonal elements are equal to one another and the remaining elements are equal to one another. The matrix calculating circuit includes a register for successively latching "n" items of data that are time-sequentially inputted thereto, a delay circuit for delaying the data supplied from the register by "n" clocks, a total-sum calculating unit for calculating a total sum of the "n" items of data supplied from the register, a data latch for latching a value of the total-sum calculating unit, and an adder for adding output data of the delay circuit to output data of the data latch. The "n" items of data to be latched by the register may be supplied from an output portion of an image sensor, to remove the influence due to the crosstalk.

Abstract: The disclosed method comprises a learning phase, during which the object zone to be recognized, selected in a learning image, is modelized, and a recognition phase during which the modelized zone is compared with image zones of a same size taken taken from the images to be processed.