This paper proposes an optical word indexing system based on fuzzy logic for 17th century printed documents. A pre-processing stage assures only interesting word candidates are considered. The main image-space indexing procedure builds fuzzy membership functions from oriented features extracted using Gabor filter banks. It proceeds by comparing this data with candidate word features and assigning each a similarity value. Results on a significant test set found over 85% of all word occurrences with a false positive rate of less than 20%.
展开▼