【摘要】 全局颜色直方图无法体现图像颜色的空间分布情况,为了克服这一缺陷,提出了利用投影法进行图像检索的算法。利用HS I颜色模型,把图像的色调、饱和度和亮度分量分别在水平和垂直两个方向上投影,得到6个投影直方图。色调和饱和度的投影直方图反映了颜色的空间分布,而亮度的投影直方图则体现了图像的形状。然后,使用这6个投影直方图的前三阶中心矩作为特征,计算图像间的距离。实验结果表明,该算法检索速度快,具有较高的查全率和查准率。更多还原

【Abstract】 The global color histogram cannot reflect the spacial information of image’s color. To overcome this defect, a new image retrieval algorithm based on image projection is presented. By using the HSI color model, the image’s hue, saturation, and intensity were projected onto the horizontal and vertical directions, respectively, and six projective histograms were obtained. The projective histograms of hue and saturation discribe the color distribution in space, and the projective histogram of intensity reflects the shape of the image. Then, the first three central moments of these six projective histograms were taken as features to compute the distance between images. Experimental results show that this algorithm performs quickly and has good recall rate and precision.更多还原