光照不均匀、光学显微镜的放大作用、光学衍射等使微视觉图像存在动态范围大、整体像素灰度值偏低、微细结构与背景对比度低、信噪比低等不足. 为增强图像质量,文中利用对比度设计高斯平滑卷积核的尺度优化准则以合理估计光照分量,提出了尺度优化的Retinex 增强算法,并用该算法对模拟图像和实拍的同轴光光强变化的微视觉图像进行增强. 结果表明,该增强算法在压缩动态范围内既能提高对比度,又能较好地保留图像细节信息,增强效果优于同态滤波算法.

Due to the existence of non-uniform illumination, magnification effect of optical microscope and opticaldiffraction, etc. , micro-vision images are of large dynamic range, low overall pixel gray value, low contrast of finestructure to background and low signal-to-noise ratio. In order to improve the image quality, a Retinex enhancemenfalgorithm with scale optimization is proposed. This algorithm uses the contrast to design an optimization criterion ofGauss smoother convolution kernel scale for the reasonable estimation of illumination component. The algorithm isthen applied to the enhancement of simulated images and micro-vision images generated under various coaxial lightintensities. The results show that the proposed algorithm is superior to the homomorphic filtering one because it im-proves the contrast and reserves the image details within the dynamic compressing range.