基于特征融合的图像目标显著性检测方法

Image Target Saliency Detection Method Based on Feature Fusion

摘要:

为了凸显图像中的感兴趣目标,提出了基于特征融合的图像目标显著性检测方法。首先通过提取可见光图像不同尺度空间的不同特征,利用区域协方差理论融合尺度空间之间串接的不同特征,然后结合全局核密度估计体现图像的全局显著性,实现局部和全局特征融合的图像目标显著性检测。仿真结果表明,无论主观评价,还是客观指标,新方法均优于当前流行的图像显著性检测方法。

Abstract:

In order to make the interest target prominent, a saliency detection method based on feature fusion is proposed. Firstly, by extracting the different characteristics of different scale space in the visible image, different concatenated characteristics among scale space were fused by using the theory of regional covariance; finally by combining the global kernel density estimation, the global saliency of the image was reflected. Thus the image target saliency detection was realized by fusing the global and local features. Experimental results show that, regardless of the subjective assessment, or the objective indicators, the proposed method is superior to usual saliency detection methods.