基于单尺度Retinex理论的带钢图像增强算法

针对带钢表面缺陷检测中图像偏暗的问题,提出一种基于单尺度Retinex理论的图像增强算法,该算法将预处理后的带钢图像乘以比例系数,得到亮度图像的估计值,接着对反射图像采用Gamma校正,最后将Gamma校正后的反射图像与粗估计的亮度图像进行合成,得到最终的增强图像。对算法中亮度估计用到的比例系数,以及Gamma校正系数进行了详细的分析。实验结果表明,该算法避免了单尺度Retinex算法在估计亮度图像时采取的复杂运算,并且较一般的图像增强算法能够使带钢图像中较暗的细节得以凸现,在增强效果和实时性上具有更大的优越性。

Steel-Strip Image Enhancement Algorithm Based on Single-scale Retinex Theory

Aimed at the problem of quite dark images in strip surface defects detection, an image enhance- ment algorithm based on single-scale Retinex theory is proposed. Firstly the strip image, which has been pretreated, is multiplied by a ratio of coefficient to estimate the illuminatinn image, then the reffection image is enrreeted by gam- ma correction, and the final, enhanced image is formed by a synthesis of Gamma corrected reflectance image and il- lumination image. In this paper, a detailed analysis of the selection of the brightness-estimate ratio parameter and the Gamma correction parameter is presented. Compared with general image enhancement algorithms, the experimental results show that the proposed method can avoid the complex computing taken by the single - scale Retinex algorithm in estimating the illumination images and can highlight the dark details of steel strip images, thus it offers better per- formance not only in visual enhancement but also in fast processing.