Character recognition method using statistically weighted correlation

Claim 1

A method of character recognition, comprising the steps of: 1) font training or creating a font of trained characters by: (a) acquiring an image composed of a two dimensional array of pixels; (b) locating all of the characters in the image by selectively scanning columns or rows of a predetermined area of the image and comparing each pixels intensity with a reference level to determine the first pixel of each character and recording the location (column and row coordinates) of such pixel and identifying the other pixels adjacent to the first whose intensity also exceeds the reference level and recording the upper left and lower right coordinates of a box bounding each character; (c) identifying (labeling) all located characters; (d) creating all newly trained characters by storing to memory as the initial values for their mean vectors m x and their mean-square vectors q x the observed light values of each pixel for each of the trained characters; (e) updating all previously trained characters by updating each trained characters mean vector m x and mean-square vector q x with its' most recent observation vector x according to the equations ##EQU15## wherein m x (K+1) and q x (K+1) are the newly updated mean and mean-square vectors, m x (K) and q x (K) are the previously updated mean and mean-square vectors, x k+1 is the most recent vector of observed light values for the trained character, and K is the number of previous observations for each trained character; (f) repeating steps (a) thru (e) for every image in the training set; (g) determining a weight vector w x and a weighted-mean vector m wx for every trained character in the font according to the equations ##EQU16## wherein m x and q x were defined in steps (d) and (e), σ 2 x is a variance vector, w xi is the i-th component of the weight vector, m xi is the i-th component of the mean vector, c is a scaling constant and N is the number of pixels in each trained character; (2) recognizing unknown characters by: (h) acquiring a two dimensional array of pixels; (i) locating all unknown characters in a manner described in (b); (j) computing statistically weighted correlation coefficients between all unknown characters and the trained character set using the weights and weighted-means for each trained character determined in step (g); and (k) identifying all unknown characters as those trained characters with the highest weighted correlation coefficients above a threshold.

Claim 2

A method of character recognition, comprising the steps of: 1) font training or creating a font of trained characters by: (a) acquiring an image composed of a two dimensional array of pixels; (b) locating all of the characters in the image by selectively scanning columns or rows of a predetermined area of the image and comparing each pixels intensity with a reference level to determine the first pixel of each character and recording the location (column and row coordinates) of such pixel and identifying the other pixels adjacent to the first whose intensity also exceeds the reference level and recording the upper left and lower right coordinates of a box bounding each character; (c) identifying (labeling) all located characters; (d) creating all newly trained characters by storing to memory as the initial values for their mean vectors m x and their mean-square vectors q x the observed light values of each pixel for each of the trained characters; (e) updating all previously trained characters by updating each trained characters mean vector m x and mean-square vector q x with its' most recent observation vector x according to the equations ##EQU15## wherein m x (K+1) and q x (K+1) are the newly updated mean and mean-square vectors, m x (K) and q x (K) are the previously updated mean and mean-square vectors, x k+1 is the most recent vector of observed light values for the trained character, and K is the number of previous observations for each trained character; (f) repeating steps (a) thru (e) for every image in the training set; (g) determining a weight vector w x and a weighted-mean vector m wx for every trained character in the font according to the equations ##EQU16## wherein m x and q x were defined in steps (d) and (e), σ 2 x is a variance vector, w xi is the i-th component of the weight vector, m xi is the i-th component of the mean vector, c is a scaling constant and N is the number of pixels in each trained character; (2) recognizing unknown characters by: (h) acquiring a two dimensional array of pixels; (i) locating all unknown characters in a manner described in (b); (j) computing statistically weighted correlation coefficients between all unknown characters and the trained character set using the weights and weighted-means for each trained character determined in step (g); and (k) identifying all unknown characters as those trained characters with the highest weighted correlation coefficients above a threshold.