Scene-Text Analysis

Abstract

Scene text analysis is the process of analyzing text in scene images for different applications. Scene text detection, recognition, and spotting methods have gained great significance in computer vision and multimedia research groups. Scene text analysis is widely accepted because of different real-life applications, like real-time traffic sign recognition, autonomous vehicles, blind navigation assistance, and multilingual translation. This is however a challenging task since scene text regions have a large variation of scale, aspect ratio, orientation, color, language, font, and script. Such text instances can also be in horizontal, oriented, and curved forms. Therefore, we illustrated various scene text detection, recognition, and spotting mechanism that can handle the traditional shortcomings in the scene text analysis. We have also explained the several deep network architectures that are utilized in the current scenario for scene text analysis. We also compare the detection, recognition, and spotting results for popular approaches on the publicly available scene text datasets.