method: Stradvision-22015-04-03

Authors: H. Cho, M. Sung, and B. Jun

Description: First, we preprocess the input image with image quality enhancement techniques, and extract character candidates using extremal regions (ER). Second, we verify the extracted character candidates with the character classifier trained by Agile Learning. Afterwards, we do text-patch matching which greatly enhances the recall rate, and group the characters into text regions. Then, we apply a deep neural network for character recognition. Finally, we filter the detected texts again using the recognition results as a reliable feedback.