It is becoming important to realize a cross-cultural communication environment among societies with different cultures. Images are effective media for exchanging cultural characteristics across cultures. This paper presents a culture-dependent color-emotion model for cross-culture oriented image retrieval system that realizes color-emotion spaces to search images with human emotion aspects. Many image retrieval systems have been featured by color analysis, but culture-dependent aspects of images are not considered intensively. Our system creates color impression spaces based on Ekman's 17 basic emotions. For the first step, the culture-dependent color impression space is created by using cultural-features. We apply automatic clustering using our previous method "Valley Tracing" in order to generate dynamic representative colors. This system automatically creates a set of culture-dependent color impression metadata.