钢板表面缺陷图像非均匀光照校正及分割算法

Non-uniform illumination correction and segmentation algorithm for steel plate surface defects images

针对钢板表面缺陷图像光照不均匀的问题,首先根据图像的亮度与光照分量和反射分量的相关性,采用多尺度高斯函数方法提取图像的光照分量并分离出反射分量.然后根据光照分量的分布情况调整二维(2D)伽玛函数的参数,实现光照分量分布不均的自适应校正,再将校正后的光照分量与原有的反射分量融合得到光照均匀的新图像.最后采用遗传算法选择最大熵阈值进行缺陷分割.结果表明:钢板表面缺陷图像的非均匀光照得到了有效改善,缺陷细节得到了较好保持;图像阈值分割缩短了寻找阈值的时间,能够有效检测出钢板表面的多种缺陷.

Aiming at the non-uniform illumination of the steel plate surface defects images,according to the correlation between the brightness of the image and the illumination component and the reflection component,the multi-scale Gaussian function method was used to extract the illumination component of the image and separate the reflection component firstly.Then,the parameters of the two-dimension(2D)Gamma function were adjusted according to the distribution of the illumination component to realize the adaptive correction of the non-uniform distribution of the illumination component.Furthermore,the new image with uniform illumination was obtained by fusing the corrected illumination with the original reflection component.Finally,genetic algorithm was used to select the maximum entropy threshold for defects segmentation.Results show that the non-uniform illumination of the steel plate surface defects images are effectively improved,and the defects details are well maintained;the image threshold segmentation shortens the time to find the threshold,which could effectively detect a variety of defects of the steel plate surface.