基于MATLAB的直方图和区域对比度相结合的图像显著性检测

Saliency Detection in Image with Method of Combination of Histogram and Region Contrast Based on MATLAB

在分析现有显著性检测模型的基础上,采用直方图和区域对比度相结合的方法对彩色图像进行显著性值计算。基于颜色直方图得到原彩色图像的显著图,使用聚类方法将图像分割成不同的区域,计算区域之间的相似性,把区域中心间的欧氏距离作为区域之间的空间权值,计算得到原彩色图像的另一显著图。为两幅显著图分配和为1的权值,相加得到最终的显著图。结果表明,此方法获得的显著区域比较符合人眼视觉效果。

On the basis of analyzing the existing model of saliency detection,the author considerscalculating significant values of color image with the method of combination of histogram and regioncontrast in this paper.Firstly,we get the saliency image of the original color image based on colorhistogram.Then we use the clustering method to segment the input image into different regions,calculate the similarities between the regions and weight the similarities with the Euclidean distance ofregional center.In this way,we obtain the another saliency image of the original color image.Finally,two weights with sum of one are reassigned to the two saliency images,and add them,we obtain thefinal saliency image.The results show the significant regions with the proposed method are relativelyconsistent with the effects of human visual system.