engleski

Scene Text Extraction using K-means Clustering in HSI Color Space: Influence of Color Distance Measure

Text extraction in scene text images is necessary step that enables higher recognition performance. Challenges in this task arise from wide set of degradations present in natural scene images like complex backgrounds, uneven illumination, viewing angle, etc. In this paper we investigate influence of color distance measure on scene text extraction performance of K-means algorithm in HSI color space. For comparison purposes we tested following clustering distances: hue distance, saturation distance, intensity distance, chromatic distance and cylindrical distance. Obtained results are analyzed with respect to their complementarity in order to show potential for performance improvement

scene text extraction; K-means; cylindrical distance; intensity distance

nije evidentirano