基于鲁棒性主成分分析的低照度图像增强算法

LOW-LIGHT IMAGE ENHANCEMENT ALGORITHM BASED ON ROBUST PRINCIPAL COMPONENT ANALYSIS

由于低照度图像对比度和信噪比低,传统图像增强算法在提高图像对比度的同时容易造成噪声放大。针对该问题,提出基于鲁棒性主成分分析(RPCA)的低照度图像增强算法。算法依据Retinex理论将图像分解为照度分量和反射分量,使用伽马矫正对照度分量进行增强。将增强后的照度分量与反射分量合成为最终的增强图像。其中图像分解采用RPCA方法实现,因为该方法可以有效地将照度信息与噪声分离,从而避免增强照度分量时放大噪声。为了提高计算效率,算法采用非精确增广拉格朗日乘子法(Inexect-ALM,IALM)求解RPCA分解问题。实验结果表明,该算法在增强图像对比度的同时避免了放大噪声,其主观评价与客观指标都优于几种经典的图像增强算法,有较好的视觉效果和较低的计算复杂度。

Because of the low contrast and signal-to-noise ratio of low-illuminance images,traditional image enhancement algorithms can easily cause noise amplification while improving image contrast.In view of this,a low-light image enhancement algorithm based on robust principal component analysis(RPCA)is proposed.The algorithm decomposed the image into illuminance component and reflection component based on Retinex theory,and then used gamma correct to enhance the illuminance component.The enhanced illumination component and reflection component were combined into the final enhanced image.The image decomposition was realized by the RPCA method,because this method could effectively separate the illuminance information from the noise,so as to avoid amplifying the noise when the illuminance component was enhanced.In order to improve the computational efficiency,the algorithm used the inexact augmented Lagrange multiplier(IALM)method to solve the RPCA decomposition problem.Experimental results show that the algorithm avoids amplification noise while enhancing image contrast,and its subjective evaluation and objective indicators are better than several classic image enhancement algorithms,with better visual effects and lower computational complexity.