彩色自然场景统计显著图模型

A Natural Scene Statistical Saliency Map Model for Color Images

提出了一种彩色图像自然场景统计显著图模型,它根据人类视觉系统对图像的处理方式,利用自然场景高斯尺度混合(GSM)统计分布中的乘数随机变量来计算图像灰度通道与彩色拮抗对的显著性描述,将三者的加权平均作为彩色图像的显著图.实验结果表明,彩色拮抗对通道的加入能够有效提高显著图模型与视觉注意力选择机制的一致性.对比不同模型提取的显著图,以及利用公开数据库计算得到的ROC曲线及该曲线下的面积(AUC),均表明本文显著图模型具有显著的优越性.

A natural scene statistical saliency map model for color images is proposed. Based on the Gaussian scale mixture distribution of natural scene statistics, the multiplier variable in the Gaussian scale mixture distribution is extracted to model the saliency map of different channels of natural images, based on the processing mechanism of the human visual system. The experimental results prove that the consistency with the visual attention mechanism of human visual system is improved by combining the color double-opponent channels. The comparison results with the ROC （Receiver Operating Characteristics） curves and their AUC s （Area Under the Curve） of the images from the open databases demonstrate that the proposed saliency map model is better than those reported.