线结构光焊接图像去噪方法

Denoising of line structured light welded seams image based on adaptive top-hat transform

为了滤除焊接过程中大量散射及飞溅焊渣等噪声,提出了一种基于自适应顶帽变换(adaptive Top-Hat transform)的线结构光焊接图像去噪方法.将噪声图像进行一定范围结构元尺寸的Top-Hat变换处理,提出了一种评价指标互相关系数(cross-correlation coefficient of image histogram,CCIH)选取得到最优结构元尺寸L.其次,进行一定迭代次数范围的顶帽变换(Top-Hat)变换处理,并提出一种基于最大类间方差(Otsu)改进结构相似度(SSIM)算法的评价指标相亮比(ratio of structural similarity index to average brightness,RSB)选取得到最优迭代次数I.结果表明,与自适应中值法(median filter,MF)、全变分法(total variation,TV)和非下采样轮廓波变换与全变差法相结合的方法(non-subsampled contourlet transform with total variation,NSCT-TV)相比,该方法在主观视觉效果,信息熵(entropy,EN)、峰值信噪比(peak signal-to-noise ratio,PSNR)和均方根误差(mean-square error,MSE)上均有较大的改善.噪声得到有效去除的同时图像中的线结构光区域得到较好的保留.

Due to the scattering of the welding material,surface shape and laser line,the brightness distribution of the line structure is uneven and there are a lot of scattering noise around it,which affects the subsequent feature point extraction.In order to filter out the noise of the weld image of line structured light,a denoising model of weld image based on adaptive Top-Hat transformation is proposed.First,the noise image is transformed to a certain range of structural element size by Top-hat transform,and then an evaluation index was proposed,cross-correlation coefficient of image histogram(CCIH),and the optimal structural element size L is selected.Second,on the premise of the optimal structural element size L,the noise image is processed by the top-hat transform with a certain range of iterations times,a new model based on Otsu proposed SSIM model is proposed to select the optimal iterations time I from the evaluation index,Ratio of Structural Similarity Index to Average Brightness(RSB).Experimental results show that,compared with the method of Median Filter(MF),the method of Total Variation(TV),and the method combining Non-Subsampled Contourlet Transform with Total Variation(NSCT-TV),the proposed method has a great improvement in the aspects of subjective visual effect,entropy(EN),peak signal-to-noise ratio(PSNR)and mean-square error(MSE).Noise is suppressed more effectively and the line structured light area is preserved better in the image.