- 著者
-
白仁田 和彦
林 健一郎
大坪 昭文
- 出版者
- 一般社団法人 日本機械学会
- 雑誌
- 日本機械学会論文集 C編 (ISSN:03875024)
- 巻号頁・発行日
- vol.70, no.689, pp.192-199, 2004-01-25 (Released:2011-03-04)
- 参考文献数
- 10
This paper describes a method of giving impression words for images based on colors. 160 impression words are prepared, and colors which are strongly related to each of the 160 impression words are registered into the system in advance. In this system, we adopt Lab color space which is a color coordinate system and can obtain exact color difference between pixels and extract main colors of images based on the results of clustering in Lab color space. The number of main colors of an image is determined when the sum of pixels of the main colors is larger than 85% of all the pixels in the image, and the colors with the larger number of pixels are determined in ascending order. We propose a distance between the main colors and the registered colors. In this system mCn combinations of distances between m colors registered in an impression words and n main colors are obtained. Using the mCn distances, we calculate a minimum distance and an average distance every 160 impression words. By the comparison of the minimum distance and the average distance of each impression word, the system determines suitable impression words for the image. Experimental results show the proposed method to be effective.