使用不同频率域残差和K-SVD模型的图像消噪方法  

Image denoising method by using different frequency domain residuals and K-SVD model

作　　者：尚丽[1] 周燕[1] 孙战里[2] SHANG Li;ZHOU Yan;SUN Zhanli(School of Electronic Information Engineering, Suzhou Vocational University, Suzhou 215104, China;College of Electrical Engineering and Automation, Anhui University, Hefei 230026, China)

机构地区：[1]苏州市职业大学电子信息工程学院,江苏苏州215104 [2]安徽大学电气工程与自动化学院,安徽合肥230026

出　　处：《实验技术与管理》2019年第10期139-143,共5页Experimental Technology and Management

基　　金：国家自然科学基金资助项目(61373098);2018年度江苏省第五期“333工程”培养对象资助项目

摘　　要：图像消噪前后的残差信息包含图像的高频信息.为了提高消噪图像的质量,在轮廓波变换域内,根据噪声图像及其K-SVD消噪后不同频率子带图像的残差信息,对残差子图像块再次进行K-SVD消噪,并将其结果与消噪图像的子带图像进行融合,得到包含残差信息的消噪子带图像,最后通过轮廓波逆变换得到消噪图像.仿真实验结果证明了所提出的基于不同频率子带内残差信息和K-SVD的图像消噪方法优于轮廓波变换和K-SVD模型,具有较好的消噪性能.The residual information before and after image denoising includes the high frequency information of the image. In order to improve the quality of the image denoised and according to the residual information of noise image and its sub-band images denoised by K-SVD at different frequencies, K-SVD denoising is performed on the residual sub-images again, and the result is fused with the sub-band images of the image denoised to obtain the sub-band images denoised containing residual information. Finally, the image denoised is obtained by the inverse contour wave transform. The simulation results show that the proposed image denoising method based on in-band residual information of different frequencies and K-SVD is superior to the contour wave transform and K-SVD model and has better denoising performance.

关 键 词：图像消噪 图像残差信息 轮廓波变换 K-SVD消噪模型 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]