摘要
目前图像提取方法没有对红外偏振图像进行去噪以及删除冗余数据的预处理,存在运行时间长、图像的噪声大以及图像边缘特征提取效果差的问题。提出低质量红外偏振图像热像特征提取方法,该方法首先在稀疏元素的基础上对图像进行去噪处理,并将低秩元素内的冗余数据进行删除,获取最完整干净的图像数据;其次,在多个方向中利用小波变换得出图像模的极大值,以此计算出图像灰度值,从而确定目标物体的边缘点,将边缘点进行归一化后连接,实现图像的特征提取。实验结果表明,特征提取平均时间为0.035 1 s,信噪比峰值可达到12.8 dB,特征提取精度最高为95%,可以说明所提方法的运行时间短、图像的噪声小以及图像边缘特征提取效果好。
At present, the image extraction methods do not denoise the infrared polarization image and delete the redundant data. There are some problems, such as long running time, high image noise and poor effect of image edge feature extraction. A thermal image feature extraction method for low quality infrared polarization image is proposed. Firstly, the image is denoised based on sparse elements, and the redundant data in low rank elements are deleted to obtain the most complete and clean image data;Secondly, the maximum value of image modulus is obtained by wavelet transform in multiple directions, so as to calculate the image gray value, so as to determine the edge points of the target object, normalize and connect the edge points to realize the feature extraction of the image. The experimental results show that the average feature extraction time is 0.035 1 s, the peak signal-to-noise ratio can reach 12.8 dB, and the highest feature extraction accuracy is 95%, which shows that the proposed method has short running time, low image noise and good image edge feature extraction effect.
作者
杜永生
蒿琳
石秦峰
DU Yongsheng;HAO lin;SHI Qinfeng(Jining University,Qufu Shandong 273155,China)
出处
《激光杂志》
CAS
北大核心
2022年第11期159-163,共5页
Laser Journal
基金
山东省自然科学基金(No.201813467974)。
关键词
红外偏振
特征提取
预处理
小波变换
去噪
infrared polarization
feature extraction
preprocessing
wavelet transform
denoising
