Text feature extraction of natural scenes using Gabor wavelet transformation based on scale overlapping

Abstract

The features of scene text images determine the scene text low-quality. Existing method can not handle the issue. A method is proposed in the paper which makes use the frequency and direction selectivity of Gabor function and constructs a suitable wavelet transform to extract the feature of characters. And further the improved Gabor wavelet based on scale overlapping is proposed. The scale overlap reinforces the selection of Gabor filters to strokes width. Thus, a high robustness feature is obtained. The experiments show that the method is with good performance and is able to deal effectively with low-quality texts.