基于背景显著性和前景显著性相结合的纺织印染图像显著性检测

Significance detection of textile printing and dyeing images based on background saliency and foreground saliency

文章提出一种基于图像背景显著性和前景显著性相结合、自底向上的纺织印染图像显著性检测算法。此算法首先通过图像的边界信息,从图像边框或中心的超像素中收集背景种子并计算出基于背景的显著图;其次,通过基于脉冲耦合神经网络(PCNN)的自适应阈值方法分割基于背景的显著图来收集前景种子并计算出基于前景的显著图;最后,使用文章提出的公式,将两个显著图综合起来,并使用显著扩散和高斯衰减的方法进行优化,获得效果最佳的纺织印染图像显著图。

This paper proposes a bottom-up algorithm for textile printing and dyeing image saliency detection based on the combination of background saliency and foreground saliency.Firstly,the background seeds are collected from the hyperpixels in the frame or center of the image through the edge information of the image,and the background based saliency map is calculated.Then,the foreground seeds are collected and the foreground saliency map is calculated by the adaptive threshold segmentation method based on pulse coupled neural network(PCNN).Finally,the two saliency images are synthesized by using the formula proposed in this paper.Significant diffusion and Gaussian attenuation are used to optimize to obtain the best effect of textile printing and dyeing image saliency map.