基于高层颜色语义名称的显著性检测*

为了克服图像底层特征与高层语义之间的语义鸿沟,降低自顶向下的显著性检测方法对特定物体先验的依赖,提出一种基于高层颜色语义特征的显著性检测方法。首先从彩色图像中提取结构化颜色特征并在多核学习框架下,实现对图像进行颜色命名获取像素的颜色语义名称；接着利用图像颜色语义名称分布计算高层颜色语义特征,再将其与底层的Gist特征融合,通过线性支持向量机训练生成显著性分类器,实现像素级的显著性检测。实验结果表明,本文算法能够更加准确地检测出人眼视觉关注点。且与传统的底层颜色特征相比,本文颜色语义特征能够获得更好的显著性检测结果。

Saliency detection based on high-level color semantic names

In order to overcome the semantic gap between low-level features and high-level semantic concepts of images and reduce the dependence of top-down saliency detection methods on particular object prior, this paper proposed a novel saliency detection method basing on high-level semantic color names. It first extracted structural color features from color images and achieved color naming of image pixels in multiple kernel learning(MKL) framework. Then high-level semantic feature maps for saliency detection were computed based on the distribution of color names. Finally, high-level semantic features and low-level Gist features were fused and imported into a linear support vector machine(SVM) to train a saliency classifier for detecting the salient or non-salient of pixels. Experimental results show that the proposed method can detect eye fixation more accurately. And comparing to traditional low-level color features, the high-level sematic color feature could obtain better saliency detection performance.