Authors: Lu Shijian, Tian Shangxuan, Lim Joo Hwee, Tan Chew Lim

Description: This method makes use of perceptual saliency of texts in scenes/born digital images. In particular, four text-specific saliency features are designed that are consistently accompanied with texts in scene/web images. Combined with the text layout information, the saliency features are integrated to classify text and non-text objects accurately. After text localization, text boundary is extracted to get the binary text images.