基于粒空间融合的多特征显著区域检测

Multi-featured saliency region detection based on granular space fusion

为更好获得图像传达的信息,提出基于粒空间融合的多特征显著区域检测方法。将图像划分为重叠的非均等矩形粒,提取每个中心粒的颜色特征;以像素点颜色值为中心,0为半径,划分图像为球形粒,对比每个粒与图像四顶点球形粒的颜色距离;利用粒之间的颜色和空间距离,提取两种粒的对比连通显著图并将之融合,获得最终显著图。实验结果表明,它更好保留了图像边缘信息,高亮显示显著区域,更有助于图像的进一步处理。

To obtain the information conveyed by the image better,a salient region detection method based on granular space was proposed. The image was divided into overlapping non-uniform rectangular granules,and the color features of each central particle were extracted. With the pixel ’s color value as the center and 0 as the radius,the image was divided into some spherical particles,and the color distance between each particle and the four vertex spherical particles of the image was compared. Using the color and spatial distance between the particles,the contrast saliency maps of the two kinds of particles were extracted and fused to obtain the final saliency map. Experimental results show that using the proposed method retains the image edge information as much as possible while highlighting the salient region,which is more helpful for further processing of the image.