Word recognition device, word recognition method, and storage medium

A word recognition method, comprising:

• reducing a capacity of a character feature dictionary used for synthesizing a word feature, comprising:
• clustering similar column or row features for respective features of a column or a row partitioned into meshes, and assigning an identification number to a clustered column or row feature;
• and
• holding the feature to which the identification number is assigned;
• extracting a feature amount of a word of an input image;
• synthesizing a feature amount of each word by synthesizing features of characters structuring each word to be recognized with the use of a character feature dictionary a capacity of which is reduced by collecting similar column or row vectors into one vector that is made representative, when the feature amount of each word to be recognized is obtained based on a word list;
• and
• making a comparison between the synthesized word feature amount and an extracted word feature amount.

A storage medium on which is recorded a program for causing a computer to execute a word recognition process, the process comprising:

• reducing a capacity of a character feature dictionary used for synthesizing a word feature, comprising:
• clustering similar column or row features for respective features of a column or a row partitioned into meshes, and assigning an identification number to a clustered column or row feature;
• and
• holding the feature to which the identification number is assigned;
• synthesizing a word feature for a comparison from column or row features within the feature dictionary, the capacity of which is reduced, based on a word list to be recognized;
• and
• extracting a feature of an input word, and performing word recognition by making a comparison between the extracted feature of the input word and the synthesized word feature.

A word recognition device recognizing a word image, comprising:

• capacity reducing means for reducing a capacity of a character feature dictionary used for synthesizing a word feature;
• synthesizing means for synthesizing a word feature for a comparison based on a word list to be recognized from column or row features within a feature dictionary a capacity of which is reduced by said capacity reducing means;
• feature extracting means for extracting a feature of an input word;
• and
• comparing means for making a comparison between the feature of the input word, which is extracted by said feature extracting means, and a synthesized word feature, and
• wherein said capacity reducing means comprises:
• clustering means for clustering similar column or row features for respective features of a column or a row partitioned into meshes, and assigning an identification number to a clustered column or row feature, and
• holding means for holding the feature to which the identification number is assigned.