Character recognition apparatus

• US 5,197,107 A
• Filed: 03/20/1990
• Issued: 03/23/1993
• Est. Priority Date: 07/20/1988
• Status: Expired due to Term

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Fukuyama, Noriyuki, Katsuyama, Yutaka, Edanami, Takafumi, Iwaki, Hiroshi
• Primary Examiner(s): Couso, Jose
• Application Number:
  US07/474,767
• Time in Patent Office: 1,099 Days
• Field of Search: 382/9, 382/30, 382/34, 382/48
• US Class Current: 382/209
• CPC Class Codes: G06V 10/20 Image preprocessing
• CPC Class Codes: G06V 10/32 Normalisation of the pattern dimensions
• CPC Class Codes: G06V 10/34 Smoothing or thinning of the pattern; Morphological operations; Skeletonisation
• CPC Class Codes: G06V 10/75 Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/146 Aligning or centring of the image pick-up or image-field
• CPC Class Codes: G06V 30/166 Normalisation of pattern dimensions
• CPC Class Codes: G06V 30/168 Smoothing or thinning of the pattern; Skeletonisation
• CPC Class Codes: G06V 30/192 using simultaneous comparisons or correlations of the image signals with a plurality of references

Abstract

A character recognition apparatus for detecting a character which is represented by an image data, by extracting a characteristic quantity of the character from the image data of the character, and comparing the characteristic quantity with characteristic quantities of characters which are memorized in a dictionary. Before the above comparison, a modification of the above characteristic quantity which is extracted from the image data is carried out, where the modification corresponds to a magnification or reduction of the scale of the above image data to equalize the scale of the above image data of the character with the scales of the characters of which the characteristic quantities are memorized in the above dictionary. Further, a magnification or reduction of the width of the character image to equalize the width of the character image with the width of the characters of which the characteristic quantities are memorized in the above dictionary.

Claims

1. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary; and

characteristic quantity magnification/reduction means for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations of said image data to equalize the scale of the character included within the image data with the scale of at least one of the characters of which the characteristic quantities are memorized in said dictionary, before said comparison,

said at least one of magnification and reduction operations of the image data by said characteristic quantity magnification/reduction means including the use of one or more parameters which correspond to the at least one of magnification and reduction operations, said character recognition apparatus further comprising;

optimum parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data, where said modification corresponds to at least one of the magnification and reduction operations of said known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with characteristic quantities which are memorized in the dictionary and which correspond to said known image data,

said comparison by said character search means being carried out by obtaining a plurality of values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

said determination of the character by said character search means being carried out by obtaining a character among the characters in the dictionary which has the highest similarity to the character included within the image data based on said plurality of values, and

said character search means further determining a recognition reliability based on the similarity of the determined character,

said character recognition apparatus further comprising;

low recognition reliability determining means for determining whether a recognition reliability of a determined character is equal to or below a first threshold value; and

optimum parameter determination starting means for obtaining an optimum parameter by restarting said optimum parameter determining means when the recognition reliability of the determined character is equal to or below said first threshold value.

2. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary; and

characteristic quantity magnification/reduction means for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations of said image data to equalize the scale of the character included within the image data with the scale of at least one of the characters of which the characteristic quantities are memorized in said dictionary, before said comparison,

said characteristic quantities each being expressed by a vector quantity comprised of a plurality of components,

each of said plurality of values which quantitatively indicate degrees of similarity being a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from said image data and the characteristic quantities memorized in said dictionary, said character recognition apparatus further comprising;

small difference determining means for determining whether each of said absolute values of differences between respective corresponding ones of the plurality of components, is equal to or below a second threshold value in said comparison; and

error accumulation preventing means for replacing each absolute value among said absolute values of differences between respective corresponding ones of the plurality of components with zero before said operation of obtaining said plurality of values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value,

the second threshold value being a value commonly used for the comparisons with absolute values corresponding to all the vector components.

3. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary; and

characteristic quantity magnification/reduction means for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations of said image data to equalize the scale of the character included within the image data with the scale of at least one of the characters of which the characteristic quantities are memorized in said dictionary, before said comparison,

said comparison in said character search means being carried out by obtaining a plurality of values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

said determination of the character in said character search means being carried out by obtaining a character among the characters in the dictionary which has the highest similarity to the character included within the image data based on said plurality of values,

said character search means further determining a recognition reliability based on the similarity of the determined character,

said character recognition apparatus further comprising

erroneous recognition character determining means for determining whether the determined recognition reliability is equal to or below a third threshold value after the character is determined so that an inaccurate recognition is detected; and

erroneous recognition character indicating means for indicating that said character is determined based on the inaccurate recognition, when the determined recognition reliability is equal to or below said third threshold value.

4. A character recognition apparatus supplied with image data, comprising:

image input means for inputting an image of a document comprised of a plurality of character regions, as the image data;

character region recognizing means for recognizing one of the plurality of character regions indicating an individual character included within said image data;

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity included within the image data in said one of the plurality of character regions;

character search means for determining the individual character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary;

text displaying means for displaying said determined characters on the display apparatus in the order that the corresponding one of the plurality of character regions is located in said image of the document;

characteristic quantity magnification/reduction means for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations of said image data to equalize the scale of the individual character with the scale of at least one of the characters of which the characteristic quantities are memorized in said dictionary, before said comparison; and

text image coordinate-corresponding memorizing means for memorizing coordinates for displaying the character on said display apparatus by said text displaying means with a correspondence to the order that the corresponding character regions are located in said image of the document.

5. A character recognition apparatus supplied with image data, comprising:

successive character strings region recognizing means for recognizing a successive character strings region including a plurality of successive character strings which include a plurality of character regions, said successive character strings region being included within the image data and said plurality of character strings being printed at intervals in the character strings region, and a continuous image region;

character string region recognizing means for recognizing each of the plurality of successive character strings in said successive character strings region;

character region recognizing means for recognizing each of the plurality of character regions indicating an individual character, from said image data;

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting characteristic quantities of characters included within the image data in each of the plurality of character regions; and

character search means for determining the characters included within the image data by comparing the characteristic quantity which are extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary;

characteristic quantity modification means for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations of said image data to equalize the scale of the character included within the image data with the scale of at least one of the characters of which the characteristic quantities are memorized in said dictionary before said comparison,

wherein said successive character strings region recognizing means comprises;

X-direction /Y-direction space string region extracting means for extracting a string region consisting of successive spaces (0 data) extending over a predetermined width and over a predetermined length, in each of the X and Y directions, to provide an X-direction space string region and a Y-direction space string region;

intermediate image composing means for composing a logical multiplication of an X-direction intermediate image and a Y-direction intermediate image to provide an intermediate image, each X-direction intermediate image and Y-direction intermediate image being included within the image data, where all data in said X-direction space string region is "zero" and all other data is "one" in said X-direction intermediate image, and all data in said Y-direction space string region is "zero" and all other data is "one" in said Y-direction intermediate image;

successive data "one" region recognizing means for recognizing a successive data "one" region in the intermediate image which is obtained by said composing operation, in a manner that a label is assigned to each of successive linear regions in the intermediate image obtained by said composing operation, where each of the plurality of successive character string regions in each of the plurality of groups contains data "one" only, extends over a predetermined width and over a predetermined length in the X-direction or the Y-direction; and

character string region determining means for projecting said image data in the direction of the character strings in a region of the image data corresponding to said successive data "one" region, and for determining a part of the image data corresponding to the projected image as the character string region when the width of the projected image is equal to or less than a predetermined width.

6. A character recognition apparatus according to claim 5,

wherein said at least one of magnification and reduction operations of the image data by said characteristic quantity magnification/reduction means includes the use of one or more parameters corresponding to the at least one of magnification and reduction operations,

wherein said comparison by said character search means is carried out by obtaining a plurality of values quantitatively indicating degrees of similarity between the characteristic quantity which is extracted from said image data, and the characteristic quantities which are memorized in said dictionary,

wherein said determination of the character by said character search means is carried out by obtaining a character among the characters in the dictionary which has the highest similarity to the characters included within the image data based on said plurality of values,

wherein a recognition reliability is determined for the character based on the determined similarity,

wherein said characteristic quantities are each expressed by a vector quantity comprised of a plurality of components,

wherein each of said plurality of values quantitatively indicating similarities are each a function of absolute values of differences between corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from the image data an the characteristic quantities memorized in the dictionary,

said character recognition apparatus further comprising:

optimum parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data, where said modification corresponds to the at least one of magnification and reduction operations of said known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data;

optimum parameter determination starting means for obtaining an optimum parameter by restarting said optimum parameter determining means when the recognition reliability of the determined character is equal to or below a first threshold value;

small difference determining means for determining whether each of said absolute values of differences between respective corresponding ones of the plurality of components, is equal to or below a second threshold value in said comparison; and

error accumulation preventing means for replacing each absolute value among said absolute values of differences between respective corresponding ones of the plurality of components with zero before said operation of obtaining said plurality of values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value.

7. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary;

character image modification means for carrying out a modification of said image data in said character region, where said modification includes at least one of thickening and thinning operations of the character included within the image data, and where said modification includes the use of one or more parameters;

characteristic quantity modification means, preceding in the next stage of said characteristic quantity extracting means, for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations which includes the use of one or more parameters;

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data,

said comparison by said character search means being carried out by obtaining a plurality of values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

said determination of the character by said character search means being carried out by obtaining a character among the characters in the dictionary which has the highest similarity to the character included within the image data based on said plurality of values,

said character search means further determining a recognition reliability based on the similarity of the determined character,

said character recognition apparatus further comprising;

low recognition reliability determining means for determining whether a recognition reliability of a determined character is equal to or below a first threshold value; and

optimum parameter determination starting means for obtaining an optimum parameter by restarting said optimum parameter determining means when the recognition reliability of the determined character is equal to or below said first threshold value.

8. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary;

character image modification means for carrying out a modification of said image data in said character region, where said modification includes at least one of thickening and thinning operations of the character included within the image data, and where said modification includes the use of one or more parameters;

characteristic quantity modification means, preceding in the next stage of said characteristic quantity extracting means, for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations which includes the use of one or more parameters;

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data,

said characteristic quantities each being expressed by a vector quantity comprised of a plurality of components,

each of said plurality of values which quantitatively indicate degrees of similarity being a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from said image data and the characteristic quantities memorized in said dictionary, said character recognition apparatus further comprising;

small difference determining means for determining whether each of said absolute values of differences between respective corresponding ones of the plurality of components is equal to or below a second threshold value in said comparison; and

error accumulation preventing means for replacing each absolute value among said absolute values of differences between respective corresponding ones of the plurality of components with zero before said operation of obtaining said plurality of values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value,

the second threshold value being a value commonly used for the comparisons with absolute values corresponding to all the vector components.

9. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary;

character image modification means for carrying out a modification of said image data in said character region, where said modification includes at least one of thickening and thinning operations of the character included within the image data, and where said modification includes the use of one or more parameters;

characteristic quantity modification means, preceding in the next stage of said characteristic quantity extracting means, for carrying out a modification of said characteristic quantity which is extracted from said image data, where said modification corresponds to at least one of magnification and reduction operations which includes the use of one or more parameters;

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data,

said comparison by said character search means being carried out by obtaining a plurality of values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

said determination of the character by said character search means being carried out by obtaining a character among the characters in the dictionary which has the highest similarity to the character included within the image data based on said plurality of values,

said character search means further determining a recognition reliability based on the similarity of the determined character,

said character recognition apparatus further comprising;

low recognition reliability determining means for determining whether a recognition reliability of a determined character is equal to or below a first threshold value; and

optimum parameter determination starting means for obtaining an optimum parameter by restarting said optimum parameter determining means when the recognition reliability of the determined character is equal to or below said first threshold value,

said characteristic quantities being each expressed by a vector quantity comprised of a plurality of components,

each of said plurality of values which quantitatively indicate degrees of similarity being a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from said image data and the characteristic quantities memorized in said dictionary, said character recognition apparatus further comprising;

small difference determining means for determining whether each of said absolute values of differences between respective corresponding ones of the plurality of components is equal to or below a second threshold value in said comparison; and

error accumulation preventing means for replacing each absolute value among said absolute values of differences between respective corresponding ones of the plurality of components with zero before said operation ob obtaining said plurality of values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value,

the second threshold value being a value commonly used for the comparisons with absolute values corresponding to all the vector components.

10. A character recognition apparatus supplied with image data, comprising:

image input means for inputting an image of a document comprised of a plurality of character regions, as the image data;

character region recognizing means for recognizing one of the plurality of character regions indicating an individual character included within said image data;

a dictionary for memorizing characteristic quantities of characters;

character image modification means for carrying out a modification of said image data in said character region, where said modification includes at least one of thickening and thinning operations of the character included within the image data, and said at least one of thickening and thinning operations is characterized by one or more parameters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within said modified image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary;

characteristic quantity modification means, for carrying out a modification of said characteristic quantity which is extracted from said image data, before said comparison, where said modification correspond to at least one of magnification and reduction operations which includes the use of one or more parameters;

text displaying means including a display apparatus which includes a plurality of coordinates, for displaying said determined characters on the display apparatus in the order that the corresponding ones of the plurality of character regions is located in said image of the document;

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data; and

text image coordinate corresponding memorizing means for memorizing coordinates for displaying said predetermined character on said display apparatus by said text displaying means with a correspondence to the order that the corresponding character regions are located in said image of the document.

11. A character recognition apparatus supplied with image data, comprising:

successive character strings region recognizing means for recognizing a successive character strings region including a plurality of successive character strings which include a plurality of character regions, said successive character strings region being included within the image data and said plurality of character strings being printed at intervals in the character strings region, and a continuous image region;

character string region recognizing mans for recognizing each of the plurality of successive character strings in said successive character strings region;

character region recognizing means for recognizing each of the plurality of character regions indicating an individual character, from said image data;

character image modification means for carrying out a modification of said image data in each of said character regions, where said modification includes at least one of magnification and reduction operations of the scale of a character and at least one of thickening and thinning operations on a character, where said at least one of magnification and reduction operations and said at least one of thickening and thinning operations include the use of one or more parameters;

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data; and

character search means for determining the character included within said image data, by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary, said character recognition apparatus further comprising;

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data,

wherein said successive character strings region recognizing means comprises;

X-direction /Y-direction space string region extracting means for extracting a string region consisting of successive spaces (0 data) extending over a predetermined width and over a redetermined length, in each of the X and Y directions, to provide an X-direction space string region and a Y-direction space string region;

intermediate image composing means for composing a logical multiplication of an X-direction intermediate image and a Y-direction intermediate image to provide an intermediate image, each X-direction intermediate image and Y-direction intermediate image being included within the image data, where all data in said X-direction space string region is "zero" and all other data is "one" in said X-direction intermediate image, and all data in said Y-direction space string region is "zero" and all other data is "one" in said Y-direction intermediate image;

successive data "one" region recognizing means for recognizing a successive data "one" region in the intermediate image which is obtained by said composing operation, in a manner that a label is assigned to each of successive linear regions in the intermediate image obtained by said composing operation, where each linear region contains data "one" only, extends over a predetermined width and over a predetermined length in the X-direction or the Y-direction; and

character string region determining means for projecting said image data in the direction of the character strings in a region of the image data corresponding to said successive data "one" region, and for determining a part of the image data corresponding to the projected image as the character string region when the width of the projected image is equal to or less than a predetermined width.

12. A character recognition apparatus according to claim 11,

wherein said comparison by said character search means is carried out by obtaining a plurality of values quantitatively indicating degrees of similarity between the characteristic quantity which is extracted from said image data, and the characteristic quantities which are memorized in said dictionary,

wherein said determination by said character among the characters in the dictionary which has the highest similarity to the character included within the image data, and

wherein said character search means further determines a recognition reliability based on the similarity of the determined character,

wherein said characteristic quantities are each expressed by a vector quantity comprised of a plurality of components,

wherein each of said plurality of values which quantitatively indicate degrees of similarity is a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from the image data and the characteristic quantities memorized in the dictionary,

said character recognition apparatus further comprising:

optimum parameter determination starting means for obtaining an optimum parameter by restarting said optimum parameter determining means when the recognition reliability of the determined character is equal to or below a first threshold value;

.

13. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

character image modification means for carrying out a modification of said image data in each of said character regions, where said modification includes at least one of magnification and reduction operations of the scale of a character and at least one of thickening and thinning operations on a character, where said at least one of magnification and reduction operations include the use of one or more parameters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within said modified image data;

character search means for determining the character included within the image data, by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary; and

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data,

said comparison by said character search means being carried out by obtaining a plurality of values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

said determination of the character by said character search means being carried out by obtaining a character among the characters in the dictionary which has the highest similarity to the character included within the image data based on said plurality of values,

said character search means further determining a recognition reliability based on the similarity of the determined character,

said character recognition apparatus further comprising;

low recognition reliability determining means for determining whether a recognition reliability of a determined character is equal to or below a first threshold value; and

optimum parameter determination starting means for obtaining an optimum parameter by restarting said optimum parameter determining means when the recognition reliability of the determined character is equal to or below the first threshold value.

14. A character recognition apparatus according to claim 13, wherein said characteristic quantities are each expressed by a vector quantity comprised of a plurality of components,

wherein each of said plurality of values which quantitatively indicate degrees of similarity is a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from said image data and the characteristic quantities memorized in said dictionary,

said character recognition apparatus further comprising:

small difference determining means for determining whether each of said absolute values of differences between respective corresponding ones of the plurality of components, is equal to or below a second threshold value in said comparison; and

error accumulation preventing means for replacing each absolute value among said absolute values of differences between respective corresponding ones of the plurality of components with zero before said operation of obtaining said plurality of values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value,

the second threshold value being a value commonly used for the comparisons with absolute values corresponding to all the vector components.

15. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

character image modification means for carrying out a modification of said image data in each of said character regions, where said modification includes at least one of magnification and reduction operations of the scale of a character and at least one of thickening and thinning operations on a character, where said at least one of magnification and reduction operations include the use of one or more parameters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within said modified image data;

character search means for determining the character included within the image data, by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary; and

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data,

said characteristic quantities each being expressed by a vector quantity comprised of a plurality of components,

each of said plurality of values which quantitatively indicate degrees of similarity being a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantities memorized in said dictionary,

said character recognition apparatus further comprising;

small difference determining means for determining whether each of said absolute values of differences between respective corresponding ones of the plurality of components, is equal to or below a second threshold value in said comparison; and

error accumulation preventing means for replacing each absolute value among said absolute values of differences between respective corresponding ones of the plurality of components with zero before said operation of obtaining said plurality of values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value,

the second threshold value being a value commonly used for the comparisons with absolute values corresponding to all the vector components.

16. A character recognition apparatus supplied with image data, comprising:

image input means for inputting an image of a document comprises of a plurality of character regions, as the image data;

character region recognizing means for recognizing one of the plurality of character regions indicating an individual character included within said image data;

a dictionary connected to memorize characteristic quantities of characters;

character image modification means for carrying out a modification of said image data in said character region, where said modification includes at least one of thickening and thinning operations of the character included within the image data, and wherein said modification includes the use of one or more parameters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the modified image data;

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary;

characteristic quantity modification means for carrying out a modification of said characteristic quantity which is extracted from said image data, before said comparison, where said modification corresponds to at least one of magnification and reduction operations which includes the use of one or more parameters;

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity which is memorized in the dictionary and which corresponds to said known image data; and

wherein said comparison in said character search means is carried out by obtaining values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

wherein said determination of the character in said character search means is carried out by obtaining a character which has the highest similarity among the characters in the dictionary based on said values,

wherein said character search means further determines a recognition reliability based on the similarity of the determined character,

said character recognition apparatus further comprising;

erroneous recognition character determining means for determining whether or not the determined recognition reliability is equal to or below a predetermined threshold value when each character is determined; and

erroneous recognition character indicating means for indicating that said character is determined based on an erroneous recognition, when the determined recognition reliability is equal to or below said predetermined threshold value.

17. A character recognition apparatus supplied with image data, comprising:

successive character strings region recognizing means for recognizing a successive character strings region including a plurality of successive character strings which include a plurality of character regions, said successive character strings region being included within the image data and said plurality of character strings being printed at intervals in the character strings region, and a continuous image region;

character string region recognizing means for recognizing each of the plurality of successive character strings in said successive character strings region;

character region recognizing means for recognizing each of the plurality of character regions indicating an individual character, from said image data;

character image modification means for carrying out a modification of said image data in each of said character regions, where said modification includes at least one of magnification and reduction operations of the scale of a character and at least one of thickening and thinning operations on a character, where said at least one of magnification and reduction operations and said at least one of thickening and thinning operations include the use of one or more parameters;

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the modified image data; and

character search means for determining the character included within said image data, by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of character which are memorized in the dictionary;

optimum modification parameter determining means for obtaining an optimum parameter of a modification by carrying out a modification of a characteristic quantity which is extracted from a known image data using each of a predetermined number of parameters, and then comparing the modified characteristic quantity with a characteristic quantity which is memorized in the dictionary and which corresponds to said known image data,

wherein said comparison in said character search mans is carried out by obtaining values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

wherein said determination of the character in said character search means is carried out by obtaining a character which has the highest similarity among the characters in the dictionary based on said values,

wherein said character search means further determines a recognition reliability based on the similarity of the determined character,

wherein said characteristic quantity is represented by a vector quantity comprised of a plurality of elements,

said character recognition apparatus further comprising;

low recognition reliability determining means for determining whether or not a recognition reliability of a determined character is equal to or below a first threshold value,

wherein each of said plurality of values which quantitatively indicate degrees of similarity is a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from said image data and the characteristic quantities memorized in said dictionary,

optimum parameter determination starting means for obtaining an optimum parameter by restarting said optimum parameter determining means when the recognition reliability of the determined character is equal to or below said first threshold value;

small difference determining means for determining whether or not each of said absolute values of differences between said corresponding components, is equal to or below a second threshold value, in said comparison; and

an error accumulation preventing means for replacing each absolute value among said absolute values of differences between said corresponding components, with zero, before said operation of obtaining said values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value;

wherein said successive character strings region recognizing means comprises;

X-direction /Y-direction space string region extracting means for extracting a string region consisting of successive spaces (0 data) extending over a predetermined width and over a predetermined length, in each of the X and Y directions, to provide an X-direction space string region and a Y-direction space string region;

intermediate image composing means for composing a logical multiplication an of an X-direction intermediate image and a Y-direction intermediate image to provide an intermediate image, each X-direction intermediate image and Y-direction intermediate image being included within the image data, where all data in said X-direction space string region is "zero" and all other data is "one" in said X-direction intermediate image, and all data in said Y-direction space string region is "zero" and all other data is "one" in said Y-direction intermediate image;

successive data "one" region recognizing means for recognizing a successive data "one" region in the intermediate image which is obtained by said composing operation, in a manner that a label is assigned to each of a plurality of successive linear regions in the intermediate image obtained by said composing operation, where each of the plurality of successive linear regions groups contains data "one" only, extends over a predetermined width and over a predetermined length in the X-direction or the Y-direction; and

and

character string region determining means for projecting said image data in the direction of the character strings in a region of the image data corresponding to said successive data "one" region, and for determining a part of the image data corresponding to the projected image as the character string region when the width of the projected image is equal to or less than a predetermined width.

18. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data; and

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary,

wherein said characteristic quantities are each expressed by a vector quantity comprised of a plurality of components, and

wherein said plurality of values which quantitatively indicate degrees of similarity is a function of absolute values of differences between respective corresponding ones of the plurality of components of the vector quantities of said characteristic quantity extracted from said image data and the characteristic quantities memorized in said dictionary,

said character recognition apparatus further comprising;

small difference determining means for determining whether each of said absolute values of differences between respective corresponding ones of the plurality of components, is equal to or below a second threshold value in said comparison; and

error accumulation preventing means for replacing each absolute value among said absolute values of differences between respective corresponding ones of the plurality of components with zero before said operation of obtaining said plurality of values which indicate degrees of similarity, when the absolute value is equal to or below said second threshold value,

the second threshold value being a value commonly used for the comparisons with absolute values corresponding to all the vector components.

19. A character recognition apparatus supplied with image data, comprising:

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data; and

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary,

wherein said characteristic quantity is expressed by a vector quantity comprised of a plurality of components, and

wherein said comparison by said character search means is carried out by obtaining a plurality of values which indicate degrees of similarity between said characteristic quantity extracted from said image data and respective characteristic quantities memorized in said dictionary,

wherein said determination of the character by said character search means is carried out by obtaining a character among the characters in the dictionary which has the highest similarity to the character included within the image data based on said plurality of values,

wherein said character search means further determines a recognition reliability based on the similarity of the determined character,

said character recognition apparatus further comprising;

an erroneous recognition character determining means for determining whether the determined recognition reliability is equal to or below a third threshold value after the character is determined so that an inaccurate recognition is detected; and

erroneous recognition character indicating means for indicating that the character is determined based on the inaccurate recognition, when the determined recognition reliability is equal to or below said third threshold value.

20. A character recognition apparatus supplied with image data, comprising:

successive character strings region recognizing means for recognizing a successive character strings region including a plurality of successive character strings which include a plurality of character regions, said successive character strings region being included within the image data and said plurality of character strings being printed at intervals in the character strings region, and a continuous image region;

character string region recognizing means for recognizing each of the plurality of successive character strings in said successive character strings region;

character region recognizing means for recognizing each of the plurality of character regions indicating an individual character, from said image data;

a dictionary for memorizing characteristic quantities of characters;

characteristic quantity extracting means for extracting a characteristic quantity of a character included within the image data in each of the plurality of character regions; and

character search means for determining the character included within the image data by comparing the characteristic quantity which is extracted from the image data with the characteristic quantities of characters which are memorized in the dictionary,

wherein said successive character strings region recognizing means comprises;

X-direction /Y-direction space string region extracting means for extracting a string region consisting of successive spaces (0 data) extending over a predetermined width and over a predetermined length, in each of the X and Y directions, to provide an X-direction space string region and a Y-direction space string region;

intermediate image composing means for composing a logical multiplication of an X-direction intermediate image and a Y-direction intermediate image to provide an intermediate image, each X-direction intermediate image and Y-direction intermediate image being included within the image data, where all data in said X-direction space string region is "zero" and all other data is "one" in said X-direction intermediate image, and all data in said Y-direction space string region is "zero" and all other data is "one" in said Y-direction intermediate image;

successive data "one" region recognizing means for recognizing a successive data "one" region in the intermediate image which is obtained by said composing operation, in a manner that a label is assigned to a linear region in the intermediate image obtained by said composing operation, and another linear region in the intermediate image, and is in contact with the linear region to which said label is assigned; and

character string region determining means for projecting said image data in the direction of the character strings in a region of the image data corresponding to said successive data "one" region, and for determining a part of the image data corresponding to the projected image as the character string region when the width of the projected image is equal to or less than a predetermined width.

21. A character recognition apparatus according to claim 20, further comprising:

erroneous recognition character determining means for determining whether the determined recognition reliability is equal to or below a third threshold value after the character is determined so that an inaccurate recognition is detected; and

erroneous recognition character indicting means for indicating that said character is determined based on the inaccurate recognition, when the determined recognition reliability is equal to or below said third threshold value.