Script Identification using Bag-of-Words with Entropy-weighted Patches

抄録

We tackle the problem of script identification, which is necessary to correctly read text in images in multilingual environments, using information entropy to prioritize discriminative patterns.