基于颜色视觉关注度的图像检索

在视觉形成的过程中,视觉、视知觉系统对不同颜色显示出不同的分辨力和关注度。在图像检索算法中,如何反应、利用这种视觉特性具有重要的研究意义。首先,根据视觉生理学和视觉心理学理论阐明了颜色关注度差异化的形成机理;然后依据Stevens法则和HSV颜色空间的六棱锥模型,提出了一种构造颜色关注度函数的方法:通过提升视觉显著度高的颜色的相应特征的关注度权值,补偿由于感知识别系统的颜色敏感度随颜色变化而引起的视觉差异。对比实验证明该文算法切实有效。

A Visual Color Attention Based Approach for Image Retrieval

During the formation of a vision,the visual system tends to show different discriminating power and attention in areas with different hue,saturation and luminance.But,how to utilize this visual phenomenon to increase the retrieval efficiency is seldom discussed.In this paper,the formation of the phenomenon is discussed in detail by visual-physiological and visual-psychological theories.Then,in light of Stevens's power law and hexagonal-pyramid model of HSV color space,a color attention based function is proposed,in which the imbalance among diverse colors caused by the visual color sensitivity is compensated by increasing the power of the relatively prominent colors.Contrast experiment proves the effectiveness of the method.