Scene Text Segmentation Method Based on MSER and MLBP

An effective algorithm for segmentation of scene text based on maximally stable extremal regions (MSER) and MLBP (Multiple Local Binary Patterns) is proposed to overcome the interference of uneven illumination and clutter background to scene text segmentation. Firstly, MSER algorithm is used to extract character candidates. Secondly, in the process of character classification, character candidates represented by the effective texture feature MLBP are verified using an AdaBoost trained classifier. Then, we use some heuristic rules to carry on character refinement. The final text segmentation output is obtained by combining the results from the R, G, B color channels in two polarities (bright text on dark background and dark text on bright background). The proposed method is evaluated on the ICDAR_2013 datasets and experiments show that it performs well and can achieve good segmentation results especially in case of uneven light and complex background.