基于邻域评价法的带钢表面缺陷图像脉冲噪声去除

Impulse noise reduction in surface defect of steel strip images based on neighborhood evaluation

针对冷轧带钢表面缺陷图像脉冲噪声的去除,通过对一些经典脉冲噪声滤波器的研究,提出了一种基于局部相似度分析和邻域噪声评价相结合的图像去噪方法。该滤波器算法在脉冲噪声评价过程中发现图像的轮廓和噪声,只处理噪声点而保持轮廓像素点不变,更有效地改善了定位噪声和保护细节的精确性。大量的仿真实验表明,这种新型的图像脉冲噪声滤波器比其他经典滤波器不论是在视觉效果上,还是在定量的评价指标上都有更好的效果,具有一定的应用价值。

A new filter is proposed based on local similarity analysis and neighborhood evaluation for denoising impulse noise of surface defect of cold-rolled steel strip images by the research of some classic impulse reduction filters. The new filter detects outliers and noise presented in the image, processes noise pixels only while keeps the outliers through a novel impulse noise evaluation process. The filter proposed significantly improves the accuracy of noise detection and detail preservation. Extensive simulation experiments indicate that the new filter outperforms other priorart methods in suppressing impulse noise in terms of perceptual visual quality and objective distortion measures. The novel filter proves to be useful in practice.