| 摘要 |
An appropriate search is carried out even with images including a complicated layout structure, decorated characters, and so on. An image search device 10 is provided with an image database 11 to store an image as a search target, a character string region extraction unit 13 to extract a character string region including a character string in the image, a character candidate recognition unit 14 to specify a plurality of character candidates through execution of character recognition from the image, for each of characters forming the character string in the character string region, a character candidate storage unit 15 to store the plurality of character candidates in the sequence of the character string in correspondence with the image as the specifying origin of the character candidates, a search keyword input unit 17 to input a search keyword, a search unit 18 to perform a search to determine whether each of characters forming the search keyword matches any of the plurality of character candidates for the character string, and an output unit 19 to output the result of the search. |
| 主权项 |
1. An information processing device comprising:
an image database for storing an image as a search target; character string region extraction means for extracting a character string region including a character string in the image stored in the image database; character candidate recognition means for specifying a plurality of character candidates through execution of character recognition from the image, for each of characters forming the character string in the character string region extracted by the character string region extraction means, wherein the plurality of character candidates are obtained for each character of the character string; character candidate storage means for storing the plurality of character candidates specified by the character candidate recognition means, for each of the characters, in correspondence with the image as the specifying origin of the character candidates; search keyword input means for inputting a search keyword; search means for performing a search to determine whether each of characters forming the keyword input by the keyword input means matches any of the plurality of character candidates obtained for the each character of the character string stored by the character candidate storage means, in a sequence of the keyword; and output means for outputting a result of the search by the search means, based on the correspondence between the character candidates and the image stored by the character candidate storage means, wherein the character candidate recognition means evaluates correctness of the character recognition on each of the character candidates specified in the character recognition and ranks each of the character candidates obtained for the same character based on the evaluated correctness, wherein the character candidate storage means stores the character candidates, based on information indicative of the correctness evaluated by the character candidate recognition means, wherein the search means is configured so that when each of the characters forming the keyword matches any of the plurality of character candidates stored by the character candidate storage means, the search means evaluates reliability on the match based on a ranking of the matched character candidate, and determines a number of character candidates to be compared with each character of the keyword according to a number of the characters of the keyword, and determines character candidates to be judged on the match with the keyword having the determined number of character candidates, from the information indicative of the correctness on the character candidates, and wherein the output means outputs the result of the search, also based on the reliability. |
