TEXT LOCATION IN COLOR SCENE IMAGES FOR INFORMATION ACQUISITION BY MOBILE TERMINALS

摘要

A camera integrated to a mobile phone for transmitting and receiving information is an objective people expect in the near future. Because texts always contain useful information, it is valuable to extract text or characters from natural pictures. This paper proposes a connected-component-based approach to automatic text location and recognition in color scene images that are taken by a digital camera. A multi-group decomposition scheme is used to deal with the complexity of the color background. Introduction of weak color and grayscale besides hue space improves the performance of the method. Block adjacency graph (BAG) algorithm is employed for extracting connected components in each image layer and alignment analysis is efficient to obtain accurate location. Some new features are applied in block candidate verification. Results of our experiments prove the efficiency for a wide range of real mobile application environments in the terms of character fonts, shooting conditions, and color backgrounds.