基于噪声检测的总变分去噪算法

对受高斯和脉冲混合噪声污染的数字图像去噪方法进行了研究,提出了一种基于噪声检测的自适应总变分(TV)去噪算法。提出的改进算法采用两步迭代框架实现:脉冲噪点检测和全变分图像恢复。第一步中,考虑到脉冲噪声污染的像素点不包含原图像有效信息,采用一种局部统计值,即邻域像素间的随机绝对差排序值(ROAD)估计出噪点的位置;第二步中,采用L2-TV方法进行去噪处理,并对上述过程进行迭代处理,得到去噪图像。在噪点估计过程中引入脉冲噪点水平参数,这样处理的优势在于可更准确地检测出脉冲噪点;而L2-TV去噪方法可很好地去除高斯噪声,两者结合有效地解决了TV算法存在误判图像脉冲噪声为边缘而产生假边缘的问题。与现有典型去噪方法的比较实验表明,该迭代去噪算法,即TV-ROAD算法,既能够去除混合噪声,又可以保留图像细节特征。

A noisedetection based total variation method for image denoising

An adaptive Total Variation (TV)denoisingmethod based on noise detectionis proposed to restore the natural images corrupted by both additive Gaussian noise and random valued impulse noise. The improved image restoration method uses a two stage iterative framework: identification of pixels corrupted by impulse noise and TV based image restoration. Firstly, the local statistical value, i.e. Rank Ordered Absolute Difference(ROAD) is used to adaptively detect the pixels corrupted by impulse noise that contains no valid information. Secondly, the L2 TV method is used to restore noise images. The two steps are processed iteratively to obtain the denoisedimages. The introduction of impulse noise level parameters in the noise estimation process has the advantage of being able to detect the impulse noise more accurately. The L2 TV de noising method can remove the Gaussian noise well. The combination of the two effectively solves the problem that the TV algorithm misjudges image impulse noise as edge and results in false edge. Compared withthe state of the art methods, the proposed imagedenoising method, named TV ROAD algorithm, can achieve superior restoration results and preserve image detail features.