摘要
在机器视觉和图像处理领域中,图像去噪是一个极其重要的问题,但在消除噪声的同时也丢失了图像中的纹理边缘信息。针对这一缺点,分析了图像去噪的难点,以UINTA(unsupervised,information-theoretic,adaptive filtering)方法为基础,对其作了改进,以信号能量为准则,分别从时域和频域的角度提出了一种纹理特征检测算子,利用该算子对滤除的残余图像重新识别,提取出被误判的纹理细节信息,然后把它补偿到滤波后的图像中,获得最终的去噪图像。实验结果表明,该方法在保留图像纹理特征的同时,有效地去除了图像中的噪声信息,提高了图像的信噪比,降低了均方误差,显著改善了图像的视觉效果,具有很强的实用性。
Image denoising is an important and widely studied problem in machine vision and image processing. However, a large number of image denoising methods eliminate noise and discard textures and edges, at the same time. To overcome the shortcomings, the paper makes its improvement on a basis of unsupervised, information-theoretic, adaptive image filter under analysis of difficulties on image denoising. According to the principles of signal energy, a detection operator of textural features is proposed to check the filter residue of the image in the time-domain and the frequency-domain respectively. Textures and details filtered out by mistake during the process of denoising will be extracted as much as possible. After the filtered image is compensated for missing information, the final denoised image is obtained. Experimental results show that the method can retain the image＇s textures and details, effectively eliminate image noise, increase the signal-to-noise ratio（SNR）, reduce the mean square error, and significantly improve the image＇s visual effect. Thus, it is practicable.
出处
《武汉大学学报（信息科学版）》
EI
CSCD
北大核心
2010年第1期102-105,共4页
Geomatics and Information Science of Wuhan University
基金
国家863计划资助项目(2006AA040307)
