Video text recognition using feature compensation as category-dependent feature extraction

Abstract

When recognizing multiple fonts, geometric features,such as the directional information of strokes, are generallyrobust against deformation but are weak against degradation.This paper describes a category-dependent feature extractionmethod that uses a feature compensation techniqueto overcome this weakness. Our proposed method estimatesthe degree of degradation of an input pattern by comparingthe input pattern and the template of each category. Thisestimation enables us to compensate the degradation in featurevalues. We apply the proposed method to the recognitionof video text suffering from degradation and deformation.Recognition experiments using characters extractedfrom videos show that the proposed method is superior tothe conventional alternatives in resisting degradation.