Pattern recognition apparatus and method

Claim 2

A pattern recognition method for scanning and saving a user-drawn image, in compressed numerical form, in a computer memory, comprising the steps of: separating shapes in said user-drawn image from a background by determining the smallest square which completely encompasses all shapes while maintaining a relative distance relationship between each shape; transforming the smallest square into a target square of predetermined size; subdividing the target square into a plurality of smaller rectangles; subdividing each rectangle of said plurality of smaller rectangles into a plurality of sectors; scanning each sector of said plurality of sectors; assigning at least one binary value to each sector, said at least one binary value being assigned to denote a first preselected graphical quality of said sector; forming a binary string by grouping together all of said binary values for each sector; forming a target list of said binary strings; forming a master list of sublists, said master list representing a map of interrelated binary strings, for a plurality of pre-scanned images; searching each of said sublists to locate a first numerical value in each sublist having a value equal to or greater than a corresponding value in said target list; preparing a list of index values where each index value represents an item in said sublists; and saving the compressed numerical form of the user-drawn image by saving said items from said target list in said sublists at positions identified by their respective index values.

Claim 3

The pattern recognition method of claim 1, wherein said at least one binary value indicates whether or not graphical data is present in a sector.

Claim 4

The pattern recognition method of claim 1, further comprising the step of assigning a second binary value to each of said sectors to indicate a second preselected graphical quality of said sector.

Claim 5

The pattern recognition method of claim 3, wherein said second binary value indicates whether or not a predetermined percentage of a sector is occupied by graphical data.

Claim 6

The pattern recognition method of claim 4, wherein said at least one binary value is given a weight greater than said second binary value.

Claim 7

A pattern recognition method for scanning and saving a user-drawn image, in compressed numerical form, in a computer memory, comprising the steps of: separating shapes in said user-drawn image from a background by determining the smallest square which completely encompasses all shapes while maintaining a relative distance relationship between each shape; transforming the smallest square into a target square of predetermined size; subdividing the target square into a plurality of smaller rectangles; subdividing each rectangle of said plurality of smaller rectangles into a plurality of sectors; scanning each sector of said plurality of sectors; assigning two binary values to each sector, a first binary value indicating whether or not graphical data is present in a sector and a second binary value indicating whether or not a sector contains at least a predetermined number of pixels that contain nonbackground data; forming a binary string by grouping together all of said binary values for each sector; forming a target list of said binary strings; forming a master list of sublists, said master list representing a map of interrelated binary strings, for a plurality of pre-scanned images; searchinq each of said sublists to locate a first item in each sublist having a value equal to or greater than a corresponding value in said target list; and preparing a list of index values where each index value represents an item in said sublists; and saving the compressed numerical form of the user-drawn image by saving said items from said target list in said sublists at positions identified by their respective index values.

Claim 8

A pattern recognition method for scanning and saving a user-drawn image, in compressed numerical form, in a computer memory, comprising the steps of: separating shapes in said user-drawn image from a background by determining the smallest square which completely encompasses all shapes while maintaining a relative distance relationship between each shape; transforming the smallest square into a target square of predetermined size; subdividing the target square into a plurality of smaller rectangles; subdividing each rectangle of said plurality of smaller rectangles into a plurality of sectors; scanning each sector of said plurality of sectors; assigning two binary values to each sector, a first binary value indicating whether or not graphical data is present in a sector and a second binary value indicating whether or not a sector contains at least a predetermined number of pixels that contain nonbackground data; forming a binary string by grouping together all of said binary values for each sector; forming a target list of said binary strings; forming a master list of sublists, said master list representing a map of interrelated binary strings, for a plurality of pre-scanned images; searchinq each of said sublists to locate a first item in each sublist having a value equal to or greater than a corresponding value in said target list; and preparing a list of index values where each index value represents an item in said sublists; and saving the compressed numerical form of the user-drawn image by saving said items from said target list in said sublists at positions identified by their respective index values.

Claim 9

A pattern recognition method, comprising the steps of: separating shapes in a target image drawn on a computer monitor screen from a background by determining the smallest square which completely encompasses all shapes while maintaining a relative distance relationship between each shape; transforming said smallest square into a target square of predetermined size; subdividing the target square into a plurality of smaller rectangles; subdividing each rectangle of said plurality of smaller rectangles into a plurality of sectors; scanning each sector of said plurality of sectors; assigning at least one binary value to each sector, said at least one binary value indicating whether or not graphical data is present in a sector; forming a binary string by grouping together all of the binary values for each sector; forming a target list of said binary strings; forming a master list of sublists, said master list representing a map of interrelated binary strings for a plurality of pre-scanned and pre-saved images; searching each of said sublists to locate a first item in each sublist having a value equal to or greater than a corresponding value in said target list; preparing a list of index values where each index value represents an item in said sublists; creating a nearness list that contains one predetermined numerical value for each image in an image database; each of said predetermined numerical values in said image database being identified by a unique image pointer; generating a nearness factor by comparing each index value from the target list with a group of values from the associated sublist in the map; saving each nearness factor in a nearness list that is indexed by means of the image pointer; sorting the nearness list so that the first item in the list is most nearly identical to the target image and so that the last item in the list is the item before the first item having a null value relative to the target image; whereby a list of image pointers and sum nearness factors is generated; whereby an image in said image database that most closely resembles the target image is identified by its image pointer; and whereby said image pointer is identified from among all other image pointers by its having a greater sum nearness factor than said other image pointers.

Claim 10

A pattern recognition method, comprising the steps of: separating shapes in a target image drawn on a computer monitor screen from a background by determining the smallest square which completely encompasses all shapes while maintaining a relative distance relationship between each shape; transforming said smallest square into a target square of predetermined size; subdividing the target square into a plurality of smaller rectangles; subdividing each rectangle of said plurality of smaller rectangles into a plurality of sectors; scanning each sector of said plurality of sectors; assigning at least one binary value to each sector, said at least one binary value indicating whether or not graphical data is present in a sector; forming a binary string by grouping together all of the binary values for each sector; forming a target list of said binary strings; forming a master list of sublists, said master list representing a map of interrelated binary strings for a plurality of pre-scanned and pre-saved images; searching each of said sublists to locate a first item in each sublist having a value equal to or greater than a corresponding value in said target list; preparing a list of index values where each index value represents an item in said sublists; creating a nearness list that contains one predetermined numerical value for each image in an image database; each of said predetermined numerical values in said image database being identified by a unique image pointer; generating a nearness factor by comparing each index value from the target list with a group of values from the associated sublist in the map; saving each nearness factor in a nearness list that is indexed by means of the image pointer; sorting the nearness list so that the first item in the list is most nearly identical to the target image and so that the last item in the list is the item before the first item having a null value relative to the target image; whereby a list of image pointers and sum nearness factors is generated; whereby an image in said image database that most closely resembles the target image is identified by its image pointer; and whereby said image pointer is identified from among all other image pointers by its having a greater sum nearness factor than said other image pointers.

Claim 11

The pattern recognition method of claim 7, wherein said at least one binary value indicates whether or not graphical data is present in a sector.

Claim 12

The pattern recognition method of claim 7, further comprising the step of assigning a second binary value to each of said sectors to indicate a second preselected graphical quality of said sector.

Claim 13

The pattern recognition method of claim 9, wherein said second binary value indicates whether or not a predetermined percentage of a sector is occupied by graphical data.

Claim 14

The pattern recognition method of claim 10, wherein said at least one binary value is given a weight greater than said second binary value.

Claim 15

A pattern recoqnition method, comprising the steps of: separating shapes in a target image drawn on a computer monitor screen from a background by determining the smallest square which completely encompasses all shapes while maintaining a relative distance relationship between each shape; transforming said smallest square into a target square of predetermined size; subdividing the target square into a plurality of smaller rectangles; subdividing each rectangle of said plurality of smaller rectangles into a plurality of sectors; scanning each sector of said plurality of sectors; assigning two binary values to each sector, a first binary value indicating whether or not graphical data is present in a sector and a second binary value indicating whether or not a sector contains at least a predetermined number of pixels that contain nonbackground data; forming a binary string by grouping together all of the binary values for each sector; forming a target list of said binary strings; forming a master list of sublists, said master list representing a map of interrelated binary strings for a plurality of pre-scanned and pre-saved images; searching each of said sublists to locate a first item in each sublist having a value equal to or greater than a corresponding value in said target image; preparing a list of index values where each index value represents an item in said sublists; creating a nearness list that contains one predetermined numerical value for each image in an image database; each of said predetermined numerical values in said image database being identified by a unique image pointer; generating a nearness factor by comparing each value from the target list with a group of values from the associated sublist in the map; saving each nearness factor in a nearness list that is indexed by means of the image pointer; sorting the nearness list so that the first item in the list is most nearly identical to the target image and so that the last item in the list is the item before the first item having a null value relative to the target image; whereby a list of image pointers and sum nearness factors is generated; whereby an image in said image database that most closely resembles the target image is identified by its image pointer; and whereby said image pointer is identified from among all other image pointers by its having a greater sum nearness factor than said other image pointers.

Claim 16

A pattern recoqnition method, comprising the steps of: separating shapes in a target image drawn on a computer monitor screen from a background by determining the smallest square which completely encompasses all shapes while maintaining a relative distance relationship between each shape; transforming said smallest square into a target square of predetermined size; subdividing the target square into a plurality of smaller rectangles; subdividing each rectangle of said plurality of smaller rectangles into a plurality of sectors; scanning each sector of said plurality of sectors; assigning two binary values to each sector, a first binary value indicating whether or not graphical data is present in a sector and a second binary value indicating whether or not a sector contains at least a predetermined number of pixels that contain nonbackground data; forming a binary string by grouping together all of the binary values for each sector; forming a target list of said binary strings; forming a master list of sublists, said master list representing a map of interrelated binary strings for a plurality of pre-scanned and pre-saved images; searching each of said sublists to locate a first item in each sublist having a value equal to or greater than a corresponding value in said target image; preparing a list of index values where each index value represents an item in said sublists; creating a nearness list that contains one predetermined numerical value for each image in an image database; each of said predetermined numerical values in said image database being identified by a unique image pointer; generating a nearness factor by comparing each value from the target list with a group of values from the associated sublist in the map; saving each nearness factor in a nearness list that is indexed by means of the image pointer; sorting the nearness list so that the first item in the list is most nearly identical to the target image and so that the last item in the list is the item before the first item having a null value relative to the target image; whereby a list of image pointers and sum nearness factors is generated; whereby an image in said image database that most closely resembles the target image is identified by its image pointer; and whereby said image pointer is identified from among all other image pointers by its having a greater sum nearness factor than said other image pointers.