基于多尺度Retinex和NSCT的泡沫图像增强方法

Bubble image enhancement based on multi-scale Retinex and NSCT

针对矿物浮选过程中获取的泡沫图像易受环境光照影响、噪声干扰和存在灰度对比度低等问题,提出一种结合多尺度Retinex(MSR)算法和非下采样Contoudet变换(NSCT)的泡沫图像增强方法。该方法首先针对光照使得浮选泡沫图像存在亮度不均,用一种区域自适应分割的MSR算法,通过调整权值改善图像的整体亮度均匀性;然后采用NSCT,通过构造分类函数完成对包含细节和噪声的高频系数处理,有效地弥补了Retinex算法在细节增强效果和噪声消除方面的不足。实验仿真结果表明,该方法能有效增强泡沫图像的轮廓、边缘和细节,抑制噪声,明显改善泡沫图像的视觉效果,为浮选泡沫图像的特征提取和品位分析奠定基础。

To solve the problem m mmera! tlotahon that the obtained bubble images are easily miluenced by lllummatmn and noises and have the low contrast, an improved image enhancement method based on the multi-scale Retinex （MSR） algorithm and the nonsubsampled Contourlet transform （NSCT） was proposed. The method uses the MSR algorithm for regional adaptive segmentation to enhance the overall brightness uniformity through adjusting the image weight, and then uses the NSCT to process the high frequency coefficients of details and noises by constructing a classification function, aiming to effectively remove the noise and enhance the weak edge, which the Retinex algo- rithm fails to do. The experimental results confirm the method＇ s positive effects in enhancing image contour, edge and details, curbing noises and improving the overall visual effect to bubble images, which lays a foundation for im- age feature extraction and mineral grade analysis.