摘要
图像显著性检测在目标识别、目标跟踪、视觉信息挖掘等研究中具有重要价值,而水下图像研究又是海洋相关学科的基础。文章针对水下图像特性,提出一种结合Retinex图像增强和超像素分割算法的多尺度显著性区域检测方法,以获取均匀、清晰的显著图。在每个尺度上进行超像素显著性估计和贝叶斯概率估计,将不同尺度的显著图进行加权求和与导向滤波,得到平滑且边缘清晰的显著图。根据水下不同倍数的衰减距离建立数据集,验证了该算法具有较强的鲁棒性。
Image saliency detection has important value in the research of target recognition, target tracking, visuaon,and the underwater image research is the basis of marine-related disciplines. According to the characteristics of underwater image,this pa-per proposes a saliency region detection algorithm to obtain uniform and clear saliency map combined witli Retinex image enhancement and multi-scale analysis on superpixels. Superpixel saliency and Bayesian probability estimation are performed on each scale,the final smooth and sharp saliency map is optimized by weighted summation and a guided filter. According to the diferent multip）les of underwater attenuation dis-tance ,datasets are established,which verifies the robustness of the algorithm.
作者
刘晓阳
薛纯
Liu Xiaoy ang Xue Chun(College of Information Science and Engineering, Ocean University of China, Qingdao 266100, China)
出处
《微型机与应用》
2017年第9期45-48,共4页
Microcomputer & Its Applications
关键词
图像增强
超像素
贝叶斯估计
显著性
image enhancement
super-pixel
Bayesian estimation
saliency detection
