基于视觉注意的图像分割方法.pdfVIP

基于视觉注意的图像分割方法 摘 要 本文提出了一种基于视觉注意的图像分割方法,该方法主要是将基于视觉注 意的显著图和边缘检测两者结合起来对图像进行分割。 首先，本文简单介绍了当前一些常用的图像分割方法，然后指出基于视觉注 意的图像分割更能符合人类的视觉习惯。 其次，介绍和分析了几种基于视觉注意的显著性检测方法，然后指出目前大 多数显著性检测方法最终只生成一幅显著图所存在的缺陷，以及仅用一幅显著图 进行图像分割时存在难以仅用简单的分割方法就能分割出与实际目标相一致的 目标区域的问题。然后指出本文的采用的图像分割方法是基于两幅显著图的分割 方法。 最后，详细介绍本文提出的图像分割方法。该方法首先生成分别基于颜色特 征和亮度特征的两幅显著图；其次，基于得到的两幅显著图，采用sobel 算子计 算得到两幅边缘图，然后合并这两幅边缘图并进行形态学操作，得到初始分割图； 最后，将初始分割图与两幅显著图相结合，去除颜色和亮度均不显著的区域，从 而得到最终的分割图。 实验表明，本文采用的图像分割方法操作简单，易于实现，速度快，而且分 割出的目标区域与实际目标一致性效果好。 关键词：图像分割；显著图；边缘检测；视觉注意 Visual Attention-based Image Segmentation Method Abstract a visual attention-based image segmentation method is proposed.In this paper,the method of image segmentation is mainly based on the combination of the saliency map of visual attention-based and the edge detection. Firstly,this paper introduced some commonly used image segmentation method which is based on visual attention, and then point out that the 1 visual attention-based image segmentation is more consistent with human visual habits. Secondly，this paper introduce and analyse several saliency detection methods which based on visual attention, and then point out the drawback which many saliency detection method all has that is only one saliency map is generated, and the problems of using only simple segmentation method is difficult to segment an area which uniformly cover the whole object.and then point out the image segmentation method of this paper is based on two saliency maps. Finally, detailedly introduce the image segmentation method of this paper proposed. The method first generates the two saliency maps which respectively based on color characteristic and luminance characteristic. Secondly, respectively based on the two saliency maps, using Sobel operator obtained two edge maps, and merge