Image recognition system

Claim 5

A method in accordance with claim 2, wherein said step of representing said skeleton image of said character by said unique reference key, further comprises: computing the cumulative reference series for each given node of said character, said cumulative reference series being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes connected to said given node, including nodes having less than or equal to the number of connections to said given node, and through nodes having less than the number of connections to said given node; sorting said cumulative reference series for each given node into a weighted order according to the number of connections connected to each said nodes; constructing a connection matrix of noes and connections between said nodes, in said weighted order; and bit mapping the connection matrix of connections between said sorted nodes to form said reference key.

Claim 6

A method in accordance with claim 1, wherein said step of representing said skeleton image of said character on the basis of internal structure corresponding to a plurality of nodes, and connections between said plurality of nodes, comprises: representing said connections between said plurality of nodes in the form of a link substantially represented as a line.

Claim 7

A method in accordance with claim 1, wherein said step of representing said skeleton image of said character on the basis of internal structure corresponding to a plurality of nodes, and connections between said plurality of nodes, comprises: representing said connections between said plurality of nodes in the form of a curve substantially represented by at least one Bezier descriptor.

Claim 8

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes and connections between said given plurality of nodes, a method for screening reference images for matching with a test image comprising: creating an image of said test character; reducing said image of said test character to skeleton image; representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node; and comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.

Claim 9

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes and connections between said given plurality of nodes, a method for screening reference images for matching with a test image comprising: creating an image of said test character; reducing said image of said test character to skeleton image; representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node; and comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.

Claim 10

An image recognition system in accordance with claim 6, wherein said step of representing said skeleton image of said test character by a reference key comprises translating a weighted matrix of nodes and connections between nodes to a binary representation of said matrix.

Claim 11

A method in accordance with claim 7, wherein said step of representing said skeleton image of said test character by said reference key, further comprises: computing the cumulative reference series for each given node of said test character, said cumulative reference series being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes connected to said given node, including nods having less than or equal to the number of connections to said given node, and through nodes having less than the number of connections to said given node; sorting said cumulative reference series for each given node into a weighted order according to the number of connections connected to each said nodes; constructing a connection matrix of nodes and connections between said nodes, in said weighted order; and bit mapping the connection matrix of connections between said sorted nodes to form said reference key.

Claim 12

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys and reference series corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of reference series being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes in said character, a method for screening reference images for matching with a test image comprising: creating an image of said test character; reducing said image of said test character to skeleton image; representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node, and by a reference series for said character, said reference series for said character being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes in said character having a given number of connections between nodes in said character; comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters to match one of said reference images in said reference image library to said test image; and comparing said reference series of said test character to each of said plurality of said reference series of said reference characters to exclude ones of said reference images in said reference library form a match with said test image.

Claim 13

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys and reference series corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of reference series being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes in said character, a method for screening reference images for matching with a test image comprising: creating an image of said test character; reducing said image of said test character to skeleton image; representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node, and by a reference series for said character, said reference series for said character being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes in said character having a given number of connections between nodes in said character; comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters to match one of said reference images in said reference image library to said test image; and comparing said reference series of said test character to each of said plurality of said reference series of said reference characters to exclude ones of said reference images in said reference library form a match with said test image.

Claim 14

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of nodes and connections between said plurality of nodes, a method for searching said stored reference images for matching with a test image comprising: creating an image of said test character; reducing said image of said test character to skeleton image; representing said skeleton image of said test character on the basis of internal structure as a plurality of nodes and connections between said plurality of nodes; and computing the ratio of the number of connections between nodes in the test image to the number of connections between nodes in the reference image for at least first and second ones of said stored reference images from said reference image library to form first and second link ratios for said first and second stored references images, respectively.

Claim 15

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of nodes and connections between said plurality of nodes, a method for searching said stored reference images for matching with a test image comprising: creating an image of said test character; reducing said image of said test character to skeleton image; representing said skeleton image of said test character on the basis of internal structure as a plurality of nodes and connections between said plurality of nodes; and computing the ratio of the number of connections between nodes in the test image to the number of connections between nodes in the reference image for at least first and second ones of said stored reference images from said reference image library to form first and second link ratios for said first and second stored references images, respectively.

Claim 16

A method in accordance with claim 10, further comprising: comparing said first link ratio for said first stored reference image to said second link ratio for said second stored reference image; and determining the likelihood of a match between one of said first and second reference images and said test image on the basis of said comparison of said first and second link ratios.

Claim 21

An apparatus in accordance with claim 13, wherein said means for representing said skeleton image of said character by said reference key, further comprises: means for computing the cumulative reference series for each given node of said character, said cumulative reference series being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes connected to said given node, including nodes having less than or equal to the number of connections to said given node, and through nodes having less than the number of connections to said given node; means for sorting said cumulative reference series for each given node into a weighted order according to the number of connections connected to each said nodes; means for constructing a connection matrix of nodes and connections between said nodes, in said weighted order; and means for bit mapping the connection matrix of connections between said sorted nodes to form said reference key.

Claim 22

An apparatus in accordance with claim 12, wherein said means for representing said skeleton image of said character on the basis of internal structure corresponding to a plurality of nodes, and connections between said plurality of nodes, comprises: means for representing said connections between said plurality of nodes in the form of a link substantially represented as a line.

Claim 23

An apparatus in accordance with claim 12, wherein said means for representing said skeleton image of said character on the basis of internal structure corresponding to a plurality of nodes, and connections between said plurality of nodes, comprises: means for representing said connections between said plurality of nodes in the form of a curve substantially represented by at least one Bezier descriptor.

Claim 24

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes and connections between said given plurality of nodes, an apparatus for screening reference images for matching with a test image comprising: means for creating an image of said test character; means for reducing said image of said test character to skeleton image; means for representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node; and means for comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.

Claim 25

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes and connections between said given plurality of nodes, an apparatus for screening reference images for matching with a test image comprising: means for creating an image of said test character; means for reducing said image of said test character to skeleton image; means for representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node; and means for comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.

Claim 26

An image recognition system in accordance with claim 17, wherein said means for representing said skeleton image of said test character by a reference key comprises translating a weighted matrix of nodes and connections between nodes to a binary representation of said matrix.

Claim 27

An apparatus in accordance with claim 18, wherein said means for representing said skeleton image of said test character by said reference key, further comprises: means for computing the cumulative reference series for each given node of said test character, said cumulative reference series being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes connected to said given node, including nodes having less than or equal to the number of connections to said given node, and through nodes having less than the number of connections to said given node; means for sorting said cumulative reference series for each given node into a weighted order according to the number of connections connected to each said nodes; means for constructing a connection matrix of nodes and connections between said nodes, in said weighted order; and means for bit mapping the connection matrix of connections between said sorted nodes to form said reference key.

Claim 28

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys and reference series corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of reference series being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes in said character, an apparatus for screening reference images for matching with a test image comprising: means for creating an image of said test character; means for reducing said image of said test character to skeleton image; means for representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node, and by a reference series for said character, said reference series for said character being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes in said character having a given number of connections between nodes in said character; means for comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters to match one of said reference images in said reference image library to said test image; and means for comparing said reference series of said test character to each of said plurality of said reference series of said reference characters to exclude ones of said reference images in said reference library from a match with said test image.

Claim 29

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys and reference series corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of reference series being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes in said character, an apparatus for screening reference images for matching with a test image comprising: means for creating an image of said test character; means for reducing said image of said test character to skeleton image; means for representing said skeleton image of said test character by a reference key, said reference key being a descriptor unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to a reference series for each node of said character, said reference series for a given node of said character being a series of integers, each integer being equal to the number of nodes having a given number of connections between nodes, connected to said given node, and by a reference series for said character, said reference series for said character being a descriptor corresponding to a series of integers, each integer being equal to the number of nodes in said character having a given number of connections between nodes in said character; means for comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters to match one of said reference images in said reference image library to said test image; and means for comparing said reference series of said test character to each of said plurality of said reference series of said reference characters to exclude ones of said reference images in said reference library from a match with said test image.

Claim 30

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of nodes and connections between said plurality of nodes, an apparatus for searching said stored reference images for matching with a test image comprising: means for creating an image of said test character; means for reducing said image of said test character to skeleton image; means for representing said skeleton image of said test character on the basis of internal structure as a plurality of nodes and connections between said plurality of nodes; and means for computing the ratio of the number of connections between nodes in the test image to the number of connections between nodes in the reference image for at least first and second ones of said stored reference images from said reference image library to form first and second link ratios for said first and second stored references images, respectively.

Claim 31

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of nodes and connections between said plurality of nodes, an apparatus for searching said stored reference images for matching with a test image comprising: means for creating an image of said test character; means for reducing said image of said test character to skeleton image; means for representing said skeleton image of said test character on the basis of internal structure as a plurality of nodes and connections between said plurality of nodes; and means for computing the ratio of the number of connections between nodes in the test image to the number of connections between nodes in the reference image for at least first and second ones of said stored reference images from said reference image library to form first and second link ratios for said first and second stored references images, respectively.

Claim 32

An apparatus in accordance with claim 21, further comprising: means for comparing said first link ratio for said first stored reference image to said second link ratio for said second stored reference image; and means for determining the likelihood of a match between one of said first and second reference images and said test image on the basis of said comparison of said first and second link ratios.

Claim 33

A method in accordance with claim 3, wherein said step of sorting said cumulative reference series for said nodes into a weighted order according to the number of connections connected to each said nodes, further comprises: sorting said cumulative reference series into an order in which those cumulative reference series having the most nodes with most links are of greater weight, those cumulative reference series having the least nodes with least links are of lesser weight, and those cumulative reference series having equal number of nodes and an equal number of links are of equal weight; and further sorting said equal weight cumulative reference series into an order wherein the cumulative reference series having a link to a node with a higher cumulative reference series is of greater weight than a cumulative reference series having a link to a node with a lower cumulative reference series.

Claim 34

An apparatus in accordance with claim 14, wherein said means for sorting said cumulative reference series for said nodes into a weighted order according to the number of connections connected to each said nodes, further comprises: means for sorting said cumulative reference series into an order in which those cumulative reference series having the most nodes with most links are of greater weight, those cumulative reference series having the least nodes with least links are of lesser weight, and those cumulative reference series having equal number of nodes and an equal number of links are of equal weight; and means for further sorting said equal weight cumulative reference series into an order wherein the cumulative reference series having a link to a node with a higher cumulative reference series is of greater weight than a cumulative reference series having a link to a node with a lower cumulative reference series.

Claim 35

A method in accordance with claim 8, wherein said step of sorting said cumulative reference series for said nodes into a weighted order according to the number of connections connected to each said nodes, further comprises: sorting said cumulative reference series into an order in which those cumulative reference series having the most nodes with most links are of greater weight, those cumulative reference series having the least nodes with least links are of lesser weight, and those cumulative reference series having equal number of nodes and an equal number of links are of equal weight; and further sorting said equal weight cumulative reference series into an order wherein the cumulative reference series having a link to a node with a higher cumulative reference series is of greater weight than a cumulative reference series having a link to a node with a lower cumulative reference series.

Claim 36

An apparatus in accordance with claim 19, wherein said means for sorting said cumulative reference series for said nodes into a weighted order according to the number of connections connected to each said nodes, further comprises: means for sorting said cumulative reference series into an order in which those cumulative reference series having the most nodes with most links are of greater weight, those cumulative reference series having the least nodes with least links are of lesser weight, and those cumulative reference series having equal number of nodes and an equal number of links are of equal weight; and means for further sorting said equal weight cumulative reference series into an order wherein the cumulative reference series having a link to a node with a higher cumulative reference series is of greater weight than a cumulative reference series having a link to a node with a lower cumulative reference series.

Claim 39

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes connected by a given plurality of connections between nodes in a given manner for each reference character, a method for screening reference images for matching with a test image comprising: creating an image of said test character; ordering said plurality of nodes in a first order; reordering said plurality of noes in a second order, wherein said second order is the same for all characters having the same number of nodes connected by the same number of connections between nodes in the same manner as said test character; representing said skeleton image of said test character on the basis of the internal structure of said test character by a descriptor corresponding to a plurality of nodes and connections between said plurality of nodes of said test character, wherein said descriptor is a reference key which is unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to said second order of said plurality of nodes; and comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.

Claim 40

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes connected by a given plurality of connections between nodes in a given manner for each reference character, a method for screening reference images for matching with a test image comprising: creating an image of said test character; ordering said plurality of nodes in a first order; reordering said plurality of noes in a second order, wherein said second order is the same for all characters having the same number of nodes connected by the same number of connections between nodes in the same manner as said test character; representing said skeleton image of said test character on the basis of the internal structure of said test character by a descriptor corresponding to a plurality of nodes and connections between said plurality of nodes of said test character, wherein said descriptor is a reference key which is unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to said second order of said plurality of nodes; and comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.

Claim 43

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes connected by a given plurality of connections between nodes in a given manner for each reference character, an apparatus for screening reference images for matching with a test image comprising: means for creating an image of said test character; means for ordering said plurality of nodes in a first order; means for reordering said plurality of nodes in a second order, wherein said second order is the same for all characters having he same number of nodes connected by the same number of connections between nodes in the same manner as said test character; means for representing said skeleton image of said test character on the basis of the internal structure of said test character by a descriptor corresponding to a plurality of nodes and connections between said plurality of nodes of said test character, wherein said descriptor is a reference key which is unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to said second order of said plurality of nodes; and means for comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.

Claim 44

In an image recognition system having a reference image library including a plurality of stored reference images, each of said stored reference images represented on the basis of the internal structure of respective reference characters including a plurality of respective unique reference keys corresponding to the internal structure of said respective plurality of reference characters, each of said plurality of unique reference keys being a descriptor corresponding to a given plurality of nodes connected by a given plurality of connections between nodes in a given manner for each reference character, an apparatus for screening reference images for matching with a test image comprising: means for creating an image of said test character; means for ordering said plurality of nodes in a first order; means for reordering said plurality of nodes in a second order, wherein said second order is the same for all characters having he same number of nodes connected by the same number of connections between nodes in the same manner as said test character; means for representing said skeleton image of said test character on the basis of the internal structure of said test character by a descriptor corresponding to a plurality of nodes and connections between said plurality of nodes of said test character, wherein said descriptor is a reference key which is unique for a given plurality of nodes and connections between said given plurality of nodes, said unique reference key corresponding to said second order of said plurality of nodes; and means for comparing said unique reference key of said test character to each of said plurality of unique reference keys of said reference characters, to match one of said reference images in said reference image library to said test image.