摘要
针对红外图像存在的加性、乘性及混合噪声,采用从图像中截取图像块,再用图像块的线性结合对原图像进行去噪,总体最小二乘算法用来求解其中的系数向量,充分考虑了噪声图像中存在的不确定性,通过这组系数得到去噪后的红外图像。在对像素点空间关系权重的求解上,采用模糊核聚类算法将红外图像粗略进行聚类,归为同一类的像素点之间存在较强空间约束关系,否则认为它们之间存在较弱空间约束。通过与维纳滤波算法比较,仿真结果证明了总体最小二乘去噪算法在红外图像的视觉质量和信噪比改善两个方面的有效性。最后通过比较无噪红外图像与去噪红外图像的直方图表明总体最小二乘去噪算法的优越性。
In order to remove noise from infrared images corrupted with additive, multiplieative, and mixed noise. An image pateh from an ideal image is modeled as a linear eombinafion of image patehes from the noisy image. The total least square (TLS)algorithm was used to solve the eoeffieients, whieh takes into aeeount the uneertainties in the measured data, and with them obtain the denoised image. The fuzzy kernel eluster algorithm was present to decide the region of weights roughly. If the pixels belonged to the same elass, there was a strong relation of spaee between them. Otherwise, there was poor. The image quality of the output and the SNR demonstrated the effectiveness of the TLS algorithms to remove noise of infrared images. The end the histograms of elean image and the denoised images were compared to prove that the TLS denoise algorithm was effective in infrared images.
出处
《激光与红外》
CAS
CSCD
北大核心
2008年第9期961-964,共4页
Laser & Infrared
关键词
红外图像去噪
总体最小二乘
乘性噪声
infrared images denoise
total least squares
the multiplieative noise