Summary: | 碩士 === 南台科技大學 === 資訊工程系 === 95 === Because computer techniques are developed continually, digital data increase rapidly. Among these digital data, the data which are based on images receive more attention, besides the data based on texts.
However, how to retrieve suitable images from image database is still a very difficult problem. The image features used in an image retrieval system contain color, texture, shape, spatial relationship, etc. Among these features, color is the most important one in color images. To represent the color component in a color image, the RGB color space is often used, but the RGB color space is not suitable for the retrieval purpose when using color as image feature. The RGB values must be always translated to other color spaces which match human vision perception more, and then the useful color features can be extracted from images. These color spaces are HSI, CIE LAB, Munsell color space, etc.
In this paper, we will adopt Munsell color space and use color dictionary, The Color Names Dictionary, which was defined by ISCC (Inter-Society Color Council) and NBS (National Bureau of Standards) in book COLOR: Universal Language and Dictionary of Names to name the color in digital color image, and then we will use these color names to address the color features of our CBIR system. According to the method we proposed, we implement two image retrieval systems. In Color Name Image Retrieval System, users can choose precise color name and precise spatial location to retrieve images. Users can use general color name and arbitrary spatial location to retrieve images from Color Group Image Retrieval System. The experiment results from two image retrieval systems are both very good.
|