摘要
为了增强归一化区域形状特征提取的稳定性,提升图像匹配效果。研究基于视觉检测的红外与可见光图像区域匹配方法。利用视觉显著性检测方法,获取红外与可见光图像的视觉显著图;通过直方图均衡化与优化配比灰度级动态范围方法,增强视觉显著图;采用仿射归一化方法,提取增强视觉显著图的归一化区域形状特征,匹配区域形状特征,完成红外与可见光图像区域匹配。实验证明:该方法可有效提升图像亮度和匹配红外与可见光图像,获取清晰度更佳的图像;在图像模糊与亮度等变化情况下,该方法边缘保持度与香农熵等分析指标值均与最高值较为接近,即图像匹配效果较优;在不同视角变化角度时,该方法归一化区域形状特征提取的稳定性较佳。
In order to enhance the stability of normalized region shape feature extraction and improve the effect of image matching,the infrared and visible imageregion matching method based on visual detection is studied. The visual saliency detection method is used to obtain the visual saliency map of infrared and visible images;the visual saliency map is enhanced by histogram equalization and optimization of gray level dynamic range;the affine normalization method is used to extract the normalized region shape features of the enhanced visual saliency map, match the region shape features, and complete the region matching of infrared and visible images. Experiments show that this method can effectively improve image brightness, match infrared and visible images, and obtain images with better definition. In the case of image blur and brightness, the analysis index values such as edge retention and Shannon entropy are close to the highest value, that is, the image matching effect is better;when different viewing angles are changed, the stability of normalized region shape feature extraction is better.
作者
罗文彬
刘敏
李琳
王成德
LUO Wenbin;LIU Min;LI Lin;WANG Chengde(Hunan University of Information Technology,Changsha,410100,China;Beijing Forestry University,Beijing 10083,China;Shandong Agricultural University,Taian 271200,China)
出处
《激光杂志》
CAS
北大核心
2023年第2期186-190,共5页
Laser Journal
基金
湖南省自然科学基金项目(No.2020JJ5397)。
关键词
视觉检测
红外图像
可见光图像
区域匹配
视觉显著图
直方图
visual detection
infrared image
visible light image
region matching
visual saliency map
histogram
