Research On The Color-Based Image Retrieval

Based on color image retrieval technology research is the current imageinformation retrieval hot research topic. For the color features simple calculation andstraight, is the most basic characteristics of image, so in content-based image retrievalhas been widely used. This subject mainly aimed at the measurement of color space,the color similarity measure and extraction of image histograms of distance.First, it is a detailed introduction that image retrieval of the key technology andresearch trend on further research basing on the color. And the paper analysis severalkinds of the color space,including HSV,HSB,RGB,LAB,CMYK.Then,it mainlydiscuss on RGB and HSV of spatial distribution,the changing laws,introduces variouscolor space and quantitative methods of color.Traditional the color space,for example HSV,it must transformate between theRGB and the HSV.the process is complexity and error. This paper studies the RGBcolor space distribution, through the observation of the three-dimensional space,reducing of proportion and section of the RGB color space of cube structure. On thebase of the workong,this papar prove the RGB color spce of distribution rule is: thecube of8vertices for center around to spread by pure color, the color to the othercolors, and gradually convert the spread of8vertices shape similarity namely RGBcolor space distribution of color is continuity. But because of each color change lawof vertices are similar but not identical, this kind of variation in the functionexpressions to describe the unification. After a great deal of experiments, this paperputs forward a new color of quantitative methods namely RGB color space model ofquantitative methods of8vertices color. Experiments show that this algorithm notonly avoid the complex color space transformation, but also more accord with eyecolor vision of similarity laws, can get a judgment of the ideal effect.Finally, in view of traditional color histogram of global spatial information, in an8vertices of color, on the basis of the quantitative method was put forward based onthe distance of the target area feature extraction algorithm histogram. For the distanceof the target area to join the histogram shape information, so making a good effect ofcomparing with traditional histogram.