基于贝叶斯理论的自适应显著性检测

摘要

Traditional saliency detection based on Bayesian theory usually uses the fixed window form and with poor adaptability. This paper presents an adaptive saliency detection algorithm based on Bayesian theory that can take the different picture with different salient objects into account. Firstly, picture edge is extracted by Canny algorithm, then an adaptive window is determined by threshold algorithm, finally the paper uses Bayesian algorithm to compute saliency map. The given adaptive window can fit better the salient objects. Experimental results show that the proposed method has higher precision and better recall compared with other traditional Bayesian algorithms and classical algorithm.%针对传统基于贝叶斯的显著性算法通常采用固定窗口的形式,适应性较差的特点,提出了一种基于贝叶斯理论的自适应显著性检测算法,该算法能够考虑到不同图像显著物体大小不同.首先采用Canny算法提取图像边缘,并利用阈值算法确定图像的自适应窗口,然后采用基于贝叶斯的滑动窗口算法计算显著图.给定的自适应窗口能够更好地契合显著物体,实验结果表明相比其他传统贝叶斯算法与经典算法,该算法具有更高的精确率和更好的召回率.