Indexing wood image for retrieval based on kansei factors

According to the strong relationship between low level features in an image and human emotion, an approach to classify wood images into emotional categories(gorgeous vs. simple) is proposed in this paper. Through analyzing the wood texture feature by the viewpoint of visual perception and image analysis, some features are structured to depict the relationship between wood texture and human emotional feelings. And the method that extract the features which reflect the emotional changes of wood texture by calculating directionality and coarseness in Tamura texture quantity and coarseness by K-means cluster algorithm, and extracting the global color feature in HSV color space is proposed, and Back Propagation Neural Network is employed to map these low level features to emotional feature space. Finally, we show some experimental results.