摘要
利用当前方法对多光谱模糊图像降噪时,未对多光谱模糊图像进行增强处理,存在图像视觉效果差、主观分数低等问题。为此,提出基于机器学习的多光谱模糊图像降噪方法。首先,利用均值滤波模板增强多光谱模糊图像色彩,同时利用高斯模板增强图像细节,将两者叠加,保证图像不受失真和光晕现象等影响,保证图像以及边界的清晰度;然后,利用核主成分分析法构建图像去噪模型,将图像坐标全部投射到特征空间中;最后,采用机器学习去噪特征空间中的近似噪点,实现多光谱模糊图像降噪。实验结果表明,所提方法的图像视觉效果较好,且主观得分较高。
When using the current method to denoise the multi spectral blurred image,it does not enhance the multi spectral blurred image,which leads to the problems of poor visual effect and low subjective score. Therefore,a multispectral image denoising method based on machine learning is proposed. Firstly,the mean filter template is used to enhance the color of the multi spectral blurred image,while the Gaussian template is used to enhance the image details,and the two are superimposed to ensure that the image is not affected by the distortion and halo phenomenon,and to ensure the clarity of the image and the boundary;Then,the image denoising model is constructed by kernel principal component analysis,and all the image coordinates are projected into the feature space;Finally,machine learning is used to denoise the approximate noise in the feature space to realize the denoising of multispectral blurred image. Experimental results show that the visual effect of the proposed method is better,and the subjective score is higher.
作者
刘玉利
王克朝
刘琳
LIU Yuli;WANG Kechao;LIU Lin(Harbin University,Harbin 150086,China)
出处
《激光杂志》
CAS
北大核心
2022年第5期156-160,共5页
Laser Journal
基金
黑龙江省自然科学基金项目(No.LH2019F046)。
关键词
机器学习
核主成分分析
图像降噪
图像增强
核函数
machine learning
kernel principal component analysis
image noise reduction
image enhancement
kernel function
