摘要
针对红外图像中目标与背景对比度低,难以准确提取目标轮廓的问题,提出一种融合符号压力函数和核差的红外图像主动轮廓分割方法。首先,采用最优核差算子捕获红外图像边界信息,结合全局和局部图像将符号压力函数进行融合,通过变分水平集方法最小化能量泛函,并依据梯度下降流迭代更新水平集;其次,将融合的符号压力函数与核差算子结合,更新主动轮廓位置,使图像分割的主动轮廓尽可能贴近目标边缘;最后,采用高斯滤波器进行水平集函数演变,得到分割结果。实验结果表明,论文提出的红外图像轮廓分割模型的平均准确率为92.93%,平均召回率为93.87%,平均Dice相似系数为0.904。
To address the issue of low contrast between targets and backgrounds in infrared images,which makes it difficult to extract target contours accurately,a novel infrared image active contour segmentation method is proposed that integrates the signed pressure force and kernel difference.First,an optimal kernel difference operator was employed to capture the boundary information from the infrared images.The signed pressure force was then fused with global and local image features,minimizing the energy function through the variational level set method and iteratively updating the level set via gradient descent flow.Second,the fused signed pressure force was combined with the kernel difference operator to update the active contour position,ensuring that the active contour aligned closely with the target edges.Finally,the level set function was evolved using a Gaussian filter to obtain the segmentation results.The experimental results demonstrate that the proposed infrared image contour segmentation model achieved an average accuracy of 92.93%,average recall rate of 93.87%,and average Dice similarity coefficient of 0.904.
作者
李传真
周斌
王伟明
闫宁
赵明冬
LI Chuanzhen;ZHOU Bin;WANG Weiming;YAN Ning;ZHAO Mingdong(School of Electronics and Electrical Engineering,Zhengzhou University of Science and Technology,Zhengzhou 450064,China;Xiong'an Institute of Innovation,Chinese Academy of Sciences,Xiong'an New Area 071000,China)
出处
《红外技术》
北大核心
2026年第3期324-329,共6页
Infrared Technology
基金
河南省科技攻关项目(242102211110)
河南省高等学校重点科研项目(24A510013)
微系统技术重点实验室开放课题(6142804231002)。
关键词
红外图像
主动轮廓分割
符号压力函数
核差算子
水平集变换
infrared image
active contour segmentation
signed pressure force
kernel difference operator
level set transformation
