GENERATION OF VISUAL PATTERN CLASSES FOR VISUAL PATTERN RECOGNITION

摘要

Example systems and methods for classifying visual patterns into a plurality of classes are presented. Using reference visual patterns of known classification, at least one image or visual pattern classifier is generated, which is then employed to classify a plurality of candidate visual patterns of unknown classification. The classification scheme employed may be hierarchical or nonhierarchical. The types of visual patterns may be fonts, human faces, or any other type of visual patterns or images subject to classification.

主权项

1. A method comprising:
classifying each of a plurality of reference visual patterns to a parent class of a visual pattern classification hierarchy; classifying each of the plurality of reference visual patterns into at least one of at least one child class and an auxiliary child class of the parent class, wherein a first reference visual pattern of the plurality of reference visual patterns is classified in the auxiliary child class and is misclassified in the parent class; modifying an image classifier configured to classify each of the plurality of reference visual patterns to one of the parent class and at least one sibling class of the parent class based on the first reference visual pattern being classified to the auxiliary child class; and reclassifying the first reference visual pattern to a first sibling class of the parent class using the modified image classifier.