为了更好地检测和识别接触网图像,首先要降低图像中的噪声.目前,BM3D是针对高斯等多种噪声降噪性能较好的算法,但它自身也存在一些不足,为此,本文提出一种基于BM3D的自适应去噪新方法(简称ABM3D).该方法在假设噪声方差未知的前提下,无需人为设定滤波阈值,通过自适应估算较为准确的阈值,实现二维和三维DC T变换域的自适应滤波,得到基础估计图像,利用估算的阈值计算出较为准确的噪声方差,实现联合维纳滤波得到最终估计图像,同时简化了参数设置,尤其是对降噪效果影响较大的参数.实验结果表明:本文提出的算法是有效的,即使在噪声强度非常高的情况下,利用其得到的降噪图像也能较好地保留边缘等细节信息,更具有实用价值.%To better detect and identify the catenary image ,the noise in the images should be first reduced .Cur-rently ,the BM3D algorithm is the most efficient denoising algorithm for the Gaussian noise and other noise models ,despite some deficiencies .To compensate for the deficiencies of the BM3D algorithm and retain better visual quality for the catenary image ,a new adaptive noise reduction method based on BM3D (called ABM3D) was proposed in this paper .Under the condition of unknown noise variance ,with the elimination of artificial setting of the filter thresholds ,accurate thresholds were obtained through adaptive calculation to obtain the im-age of basic estimate by realizing filtering in the two and three dimensional DCTtransform domains .Accurate noise variance was calculated by the use of the obtained threshold for collaborative Wiener filtering to obtain the image of final estimate .At the same time ,the setting of parameters was simplified ,especially the parameters that influenced the denoising effect significantly .The experimental results proved the effectiveness and practi-cal value of the proposed method ,as the denoised images obtained by this method retained edge and other de-tails well even in the case of relatively high levels of noise .
展开▼
