Automatic threshold determination for a digital scanner

Claim 2

A computer-based method for automatically determining a threshold level for a cell of a digital image representation of a document, the digital image being divided into one or more cells where each cell has a plurality of pixels and each pixel has associated therewith a plurality of bits to represent a graylevel for that pixel, wherein the threshold level is used to convert the plurality of bits to a one-bit-per-pixel representation of a graylevel, comprising the steps of: (a) scanning the document to form the digital image representation of the document in a computer memory; (b) constructing a co-occurrence matrix for each cell indicating a quantity of each different pixel graylevel pair; (c) determining a background and a foreground peak for each cell based on said co-occurrence matrix; (d) determining features of each cell; (e) determining a center of mass for each cell based on said features; (f) determining the threshold level for each cell based on said center of mass; (g) comparing the threshold level to the graylevel of each of the plurality of pixels in each cell; and (h) setting the pixel graylevel to a first value of a one bit representation if the graylevel of the pixel is greater than the threshold level, and setting the pixel value to a second value of a one bit representation of the graylevel of the pixel is less than the threshold level, thereby converting the plurality of bits to a one-bit-per-pixel representation of the graylevel for each oft he plurality of pixels in each cell.

Claim 3

The method of claim 1, further comprising the step of repeating said steps (b) through (h) for each cell of the document to thereby convert the plurality of bits to a one-bit-per-pixel representation of the graylevel for each cell of the document.

Claim 4

The method of claim 1, wherein said step (b) comprises the steps of i. initializing said co-occurrence matrix for the cell; ii. determining pixel graylevel pairs for the cell; and iii. updating said co-occurrence matrix for each pixel graylevel pair.

Claim 5

The method of claim 3, further comprising a step after said step (b) of collapsing said co-occurrence matrix and constructing a one-dimensional histogram from said collapsed co-occurrence matrix.

Claim 6

The method of claim 1, further comprising a step after said step (b) of summing proximate elements of said co-occurrence matrix and using said sums to construct a one-dimensional histogram.

Claim 7

The method of claim 5, wherein construction of said one-dimensional histogram comprises the steps of: (i) traversing a diagonal of said co-occurrence matrix until a first non-zero element is reached; (ii) entering the value of said first non-zero element into an element of said one-dimensional histogram corresponding to a column number of said first non-zero element; (iii) if an additional non-zero element exists along said diagonal, starting with said additional non-zero element, summing said proximate elements; (iv) entering the summed value of said proximate elements into an element of said one-dimensional histogram corresponding to a column number of said next non-zero element; and (v) repeating said steps (iii) and (iv) for each said additional nonzero element.

Claim 8

The method of claim 6, further comprising the step of entering a value of zero into each element of said one-dimensional histogram that does not have a value entered therein as a result of said steps (i) through (v).

Claim 9

The method of claim 6, wherein said proximate elements are summed using an L-shaped method.

Claim 10

The method of claim 1, wherein said step (d) comprises the steps of: (i) determining an extent for each of said peaks determined in said step (c); (ii) determining a deviation of said foreground peak determined in said step (c); and (iii) determining a delta between said foreground and said background peak determined in said step (c).

Claim 11

The method of claim 9, wherein said step (ii) comprises the steps of: (1) examining said background peak to determine if a graylevel of said foreground peak is less than a specified value; (2) collapsing columns of said co-occurrence matrix horizontally based on said graylevel of said foreground peak; (3) constructing a one dimensional histogram using said collapsed co-occurrence matrix; and (4) calculating a deviation of said foreground peak using said one-dimensional histogram constructed in said step (3).

Claim 12

The method of claim 10, wherein said deviation, D, is calculated as: ##EQU9## wherein x is a column number of said one-dimensional histogram, p is a column number for the foreground peak, and f(x) is a value in column x.

Claim 13

The method of claim 1, wherein said step (e) comprises the steps of: (i) determining an influence of the cell; (ii) determining a weight of a foreground of the cell; and (iii) determining a center of mass for the cell based on said influence of the cell and on said weight.

Claim 14

An automatic threshold determination system for determining a threshold level for a cell of a digital image representation of a document, the digital image being divided into one or more cells where each cell has a plurality of pixels and each pixel has a plurality of bits to represent a graylevel of the pixel, wherein the threshold level is used to convert the plurality of bits to a one-bit-per-pixel representation of a graylevel, comprising: (a) a digitizer configured to generate the digital image; (b) a matrix generator, configured to receive the digital image from said digitizer and to generate a co-occurrence matrix and a one-dimensional histogram based on said co-occurrence matrix; (c) a peak extractor, coupled to said matrix generator, configured to receive said one-dimensional histogram and to extract peak features and peak centers of mass for each cell from said one-dimensional histogram; (d) a threshold determination module, coupled to said peak extractor, configured to determine the threshold level for each cell using said peak centers of mass and said peak features; and (e) a comparator configured to compare the threshold level to the graylevel of each of the plurality of pixels in the cell and to set the pixel graylevel to a first value of a one bit representation if the graylevel of the pixel is greater than the threshold level, and to set the pixel value to a second value of a one bit representation if the graylevel of the pixel is less than the threshold level, thereby converting the plurality of bits to a one-bit-per-pixel representation of the graylevel for each of the plurality of pixels in the cell.

Claim 15

An automatic threshold determination system for determining a threshold level for a cell of a digital image representation of a document, the digital image being divided into one or more cells where each cell has a plurality of pixels and each pixel has a plurality of bits to represent a graylevel of the pixel, wherein the threshold level is used to convert the plurality of bits to a one-bit-per-pixel representation of a graylevel, comprising: (a) a digitizer configured to generate the digital image; (b) a matrix generator, configured to receive the digital image from said digitizer and to generate a co-occurrence matrix and a one-dimensional histogram based on said co-occurrence matrix; (c) a peak extractor, coupled to said matrix generator, configured to receive said one-dimensional histogram and to extract peak features and peak centers of mass for each cell from said one-dimensional histogram; (d) a threshold determination module, coupled to said peak extractor, configured to determine the threshold level for each cell using said peak centers of mass and said peak features; and (e) a comparator configured to compare the threshold level to the graylevel of each of the plurality of pixels in the cell and to set the pixel graylevel to a first value of a one bit representation if the graylevel of the pixel is greater than the threshold level, and to set the pixel value to a second value of a one bit representation if the graylevel of the pixel is less than the threshold level, thereby converting the plurality of bits to a one-bit-per-pixel representation of the graylevel for each of the plurality of pixels in the cell.

Claim 16

The automatic threshold determination system of claim 13, wherein said digitizer is a flatbed scanner.

Claim 17

A computer-based system for automatically determining a threshold level for a cell of a digital image representation of a document, the digital image being divided into one or more cells and each cell having a plurality of pixels and each pixel having associated therewith a plurality of bits to represent a graylevel for that pixel, wherein the threshold level is used to convert the plurality of bits to a one-bit-per-pixel representation of a graylevel, comprising: first means for scanning the document to form the digital image representation of the document in a computer memory; second means for constructing a co-occurrence matrix for each cell indicating a quantity of each different pixel graylevel pair; third means, coupled to said second means, for extracting a background and a foreground peak for each cell based on said co-occurrence matrix constructed by said second means and for determining features of said peaks extracted by said third means; fourth means, coupled to said third means for determining a center of mass for each cell based on said features; fifth means, coupled to said third and fourth mesas for determining the threshold level for each cell; sixth means for comparing the threshold level to the graylevel of each of the plurality of pixels in each cell; and seventh means for setting the pixel graylevel to a first value of a one bit representation if the graylevel of the pixel is greater than the threshold level, and setting the pixel value to a second value of a one bit representation of the graylevel of the pixel is less than the threshold level, thereby converting the plurality of bits to a one-bit-per-pixel representation of the graylevel for each of the plurality of pixels in each cell.

Claim 18

A computer-based system for automatically determining a threshold level for a cell of a digital image representation of a document, the digital image being divided into one or more cells and each cell having a plurality of pixels and each pixel having associated therewith a plurality of bits to represent a graylevel for that pixel, wherein the threshold level is used to convert the plurality of bits to a one-bit-per-pixel representation of a graylevel, comprising: first means for scanning the document to form the digital image representation of the document in a computer memory; second means for constructing a co-occurrence matrix for each cell indicating a quantity of each different pixel graylevel pair; third means, coupled to said second means, for extracting a background and a foreground peak for each cell based on said co-occurrence matrix constructed by said second means and for determining features of said peaks extracted by said third means; fourth means, coupled to said third means for determining a center of mass for each cell based on said features; fifth means, coupled to said third and fourth mesas for determining the threshold level for each cell; sixth means for comparing the threshold level to the graylevel of each of the plurality of pixels in each cell; and seventh means for setting the pixel graylevel to a first value of a one bit representation if the graylevel of the pixel is greater than the threshold level, and setting the pixel value to a second value of a one bit representation of the graylevel of the pixel is less than the threshold level, thereby converting the plurality of bits to a one-bit-per-pixel representation of the graylevel for each of the plurality of pixels in each cell.

Claim 19

The computer based system of claim 15, wherein said second means further comprises means for constructing a one-dimensional histogram based on said co-occurrence matrix.