基于DCT域的遥感图像目标显著性检测算法  

Salient target detection of remote sensing image based on DCT domain

作　　者：罗雅丹 蓝如梦 谭盛辉 LUO Yadan;LAN Rumeng;TAN Shenghui(Guangxi Institute of Mapping,Nanning 530023,China)

机构地区：[1]广西壮族自治区地图院,南宁530023

出　　处：《智能计算机与应用》2023年第2期165-168,共4页Intelligent Computer and Applications

摘　　要：为进一步提升遥感图像目标检测的速度和精度,针对目标与周围环境存在一定特征差异的特点,提出一种基于离散余弦变换(DCT)域的遥感图像目标显著性检测算法。直接在DCT域中进行显著性检测,重点计算每个DCT块的自适应标准差作为DCT块之间的差异权重系数,并结合亮度、颜色和纹理特征差异矩阵生成最终结果。实验表明,该算法可以提高运行效率和检测精度。To upgrade the velocity and accuracy of remote sensing image object detection,focusing on the differences between the target and surrounding area,this paper proposes a novel method for salient target detection based on discrete cosine transform(DCT)domain.It performs saliency detection in the DCT domain directly,and calculates its adaptive standard deviation for each DCT block as difference weight coefficient between the DCT blocks.Finally,the results generated by combining the brightness,color and texture feature difference matrix.Experimental results show that it has higher efficiency,and better detection accuracy and efficiency.

关 键 词：遥感图像 目标检测 DCT域 显著性 自适应标准差 

分 类 号：TP394.41[自动化与计算机技术—计算机应用技术]