为了取得更好的显著性检测结果,针对传统的显著性检测方法易造成边界模糊以及应用中央-周边差进行图像检测时,感兴趣目标的内部纹理会破坏目标的整体性的问题,提出了一种基于纹理抑制和连续分布估计的显著性检测方法.先对图像进行双边滤波的预处理,以平滑目标以及背景区域内部的纹理扰动,保留目标与背景之间的主要边缘.再采用SLIC超像素分割算法,对图像中具有相同特征的像素进行分组,通过多维正态分布提取分割区域的特征,利用二范数Wasserstein距离计算区域相似度:结合局部显著性检测以及全局显著性检测实现目标区域的提取.实验结果表明,本文的方法能够较好地提取显著性目标区域.

The traditional methods of saliency detection tend to blur the border and internal texture of interested target,which usually destroys the integrity of the target by applying the center-surround diffe-rences.As to this problem,this paper proposes a saliency detection method based on texture inhibition and continuous distribute estimation in order to obtain better saliency detection results.This paper proposes a saliency detection method based on texture inhibition and continuous distribute estimation.Firstly,bilateral filter is used to smooth texture disturbances of the target and background region,while retaining the major edges between the target and background.Then SLIC superpixel image segmentation algorithms are used to divide the pixels into groups which have lots of pixels with the same characteristics.Then features of segment region are extracted by multidimensional normal distribution,and the similarity of feature saliency is calcula-ted by two norm Wasserstein distance.At last,saliency region can be extracted by the local and global saliency detection.The experiment result shows that the experiment method in this paper can extract saliency region effectively.The experiment results show that our method can obtain better detection result.