A Fast Logo Recognition Algorithm in Noisy Document Images
Sina Hassanzadeh, Hossein Pourghassem · 2011
Logo recognition is one of the applied aspects of graphic recognition domain. In most of document images, some diverse conditions such as noise existence, occlusion, and different scale/orientation may affect logo recognition process. In this paper, a novel approach based on spatial and structural features of logo images is proposed to overcome those problems. After normalization step which eliminates the sensitivity of the logo recognition process to different scale/orientation, a novel feature is extracted based on horizontal and vertical histograms of the logo image. Finally, KNN classifier is used to recognize logo images. Our proposed algorithm is evaluated on a standard logo dataset of Maryland University. The experimental results show the robustness of the proposed approach in the logo recognition.