Fuji Electric Co., Ltd.

Fujifacom Corporation

Fuji Electric Corporate Research and Developement Ltd.

Yasuo Hongo

Atsushi Yoshida

Yoshio Nitta

Kouichi Toyama

Masatoshi Okada

Hitoshi Ichinohe

A method for nonlinear normalization of an image, which performs pre-processing for computing the correlation between an unknown pattern and a reference pattern. A local spatial density function .rho.(X.sub.i, Y.sub.j) (i=1-I, j=1-J) is calculated from a two-dimensional pattern f(X.sub.i, Y.sub.j) which is obtained by sampling the unknown pattern at a sampling interval .gamma.. The spatial density function .rho.(X.sub.i, Y.sub.j) is obtained as the product of reciprocals of line pitches in both the X and Y directions. An x-direction cumulative function hx(X.sub.i) and a y-direction cumulative function hy(Y.sub.j) are computed by successively adding the space density function .rho.(X.sub.i, Y.sub.j). New sampling points (X.sub.i, Y.sub.j) are computed in such a fashion that new sampling intervals (.delta.i, .epsilon.j), defined as intervals between two adjacent points of the new sampling points (X.sub.i, Y.sub.j), satisfy the condition that a product between the cumulative function hx(X.sub.i) and .delta.i takes a first fixed value, and a product between the cumulative function hy(Y.sub.j) and .epsilon.j takes a second fixed value. The normalized sampled values at the new sampling points (X.sub.i, Y.sub.j) are obtained by resampling the unknown pattern or by performing a computation on the two dimensional pattern f(X.sub.i, Y.sub.j).

Apparatus and method for nonlinear normalization of image

A feature recognition method and apparatus is described where in one embodiment the digitized two-dimensional pattern or image usually of a character to be recognized is subjected first to conversion into a vector. The vector subsequently undergoes propagation and stopping operations subject to various tests so as to generate a list of unique features about the pattern being identified. Thusly, identified features are compared mathematically to a previously created feature library until a match is found. In an alternative embodiment, to increase the speed of three-dimensional, feature-extraction processing of character patterns the vector-extraction positions on a pattern are converted into flags and accommodated in a corresponding motion-flag memory. Flag values in the motion-flag memory and a stopping-flag memory are read sequentially after propagation processing, and it is determined whether there are flag values representing the existence of vectors in the flag values in the above addresses of other flag memories and addresses adjoining thereto. The flag values in corresponding addresses in a self-moving flag memory are reset, and at the same time, the flags are set in corresponding addresses in the stopping-flag memory relative to vectors with the same direction, if there are vector flag values in the corresponding addresses. The flag values in each stopping-flag memory and each motion-flag memory are converted into vector values via determination processing after propagation processing, and are output as three-dimensional features.

Character-feature extraction device

A method for determining a binary coding threshold value for a target object comprises determining reference binarization characteristic values for each reference object, storing the reference values in a memory, selecting the reference value associated with a reference object corresponding to the target object, generating in a learning procedure a histogram corresponding to the selected reference value, extracting feature data, i.e., binarization characteristic values, in a threshold control step from the histogram, and setting the binary coding threshold value to a predetermined weight mean value calculated from the binarization characteristics. 7257

Method for determining binary coding threshold value

A binarizing apparatus used for optical character readers or character-and-graphic input devices, in which a sharp binary image can be obtained even from low-contrast character lines (or line segments) by discriminating a line graphic from stains wider than the character lines or from blurs in the background thereof. Eight surrounding subwindows are disposed around a target subwindow enclosing a target pixel, the surrounding subwindows being separated from each other at a distance larger than the character line width, and each of the subwindows having 3.times.3 pixels. The average density value of the respective subwindow is calculated to thereby compare the respective density values of the target subwindow and each surrounding subwindow or compare the density values of the respective surrounding subwindows. Consequently, whether the target pixel belongs to character line or whether it does not belong thereto, is judged to thereby perform binarizing.

Binarizing apparatus

The identity of an unknown character is determined by comparing matrices representing the unknown character with matrices representing known dictionary characters and determining the deviation between matrices for the unknown and the dictionary characters. By obtaining both matrices for the character and for the background surroundings of the character, it is possible to extract small differences between local portions of the character. The dictionary pattern of each character is automatically created by 'learning' and also the features of the character at the time of judgment and the related result are totalized so that the dictionary pattern can be created from the complementary cumulative matrices obtained by such totalization. By virtue of this advantage, it is possible to asymptotically improve the recognition rate.

Character reading method

An object identification method and apparatus determines within the boundary of an object to be identified points corresponding to abrupt changes in the profile of the object. Each segment of the boundary between adjacent points is identified with primitive element data and distance map data. These data associated with each segment are collated with similar type data associated with the boundaries of reference objects and the object is identified as the reference object having the highest coincidence of collation.

Method and apparatus for object identification

A character identifying device generates a core bit matrix and a mask bit matrix for each reference character pattern scanned by a television camera or like device. The two matrices represent binary coded representation of a reference character pattern at two different threshold levels. An enlarged matrix representation and a contracted matrix representation of the same unknown character pattern are compared to the core bit matrix and the mask bit matrix, respectively, of each reference character pattern and corresponding correlation output signals are generated to unable the identification of the reference character pattern most closely correlated to the unknown character pattern.

Character identifying apparatus

In the particular embodiment of the invention described in the specification, the identity of an unknown character is determined by comparing matrices representing the unknown character with matrices representing known characters and determining the deviation beween matrices for the unknown and each known character. To obtain the matrices to be compared, a transformation chain is expressed in terms of horizontal, vertical and diagonal vectors, permitting all types of transformation chains to be used. Two-dimensional memory units having superposed shift registers are used in comparing matrices to determine minimum deviation.

Character reader

In the particular embodiment of the invention described in the specification, a pattern recognition apparatus converts the signals received from a video pattern detector into a bit matrix comprising columns and rows of character elements. The matrix is divided into a series of blocks and the pressence of horizontal or vertical strings of bits is compared with corresponding bit information relating to a set of known characters. An unknown character is identified by a minimum deviation of the compared block information from that of a corresponding known character.

Pattern recognition apparatus

A pattern recognition apparatus for identifying characters for example comprises a scanner and data extraction circuit which obtains a rectangular sized M.times.N bit pattern matrix B(i,j) representing the presence and configuration of an unknown scanned pattern in subregions (i,j) of a frame which substantially encloses the scanned pattern, wherein each bit in the pattern has a pattern present value when the scanned pattern falls at least partially within its associated subregion and a pattern absent value when the scanned pattern does not fall within its associated subregion. A processor compares the bit pattern matrix with a plurality of previously obtained bit pattern matrices B.sub.S.sup.K (i,j) for a respective plurality of known reference patterns K, and outputs for each reference pattern K a comparison quantity value D.sup.K whose amplitude represents the degree of correspondence between the bit pattern matrix B(i,j) from scanning the unknown pattern and the bit pattern matrices B.sub.S.sup.K (i,j) for each reference pattern K. The processor then selects the lowest and next lowest comparison quantity values and outputs a pattern identification signal which identifies the unknown pattern as the reference pattern corresponding to the lowest quantity value if this value is less than a first preset value, and if the difference between the lowest and next lowest comparison quantity values is greater than a second preset value.

Growing community of inventors

Tokyo, Japan

Yasuo Hongo