基于MSRCR-拉普拉斯金字塔方法的低照度图像增强

Low Illumination Image Enhancement Based on MSRCR-Laplace Pyramid Method

摘要: 针对传统Retinex图像增强算法存在的纹理细节保留差、过度增强和色调突变等不足，文中提出了一种基于MSRCR (带色彩恢复的多尺度Retinex算法)的拉普拉斯金字塔方法，用于弱光图像增强。该方法由三个重要部分组成：照度颜色校正、反射成分细节增强和线性加权融合。首先，将伽马校正后的照度加回反射中，实现色彩增强。然后，通过拉普拉斯金字塔处理反射分量来实现细节增强。最后，细节增强的图像和颜色校正的图像通过加权融合重构出增强后的输出图像。主观与客观的性能评估表明，相较于对比算法，文中所提出的方法可以更加有效地增强暗区图像的细节和全局对比度，使得输出图像具备更好的视觉效果。因此，该方法是一种有效的弱光图像增强方法，并具有一定的工程应用价值。

Abstract: To address the shortcomings of traditional Retinex image enhancement algorithms such as poor texture detail retention, over-enhancement and tonal mutation, a Laplace pyramid method based on MSRCR (Multiscale Retinex algorithm with color recovery) is proposed in the paper for low light image enhancement. The method consists of three important parts: illumination color correction, reflection component detail enhancement, and linear weighted fusion. First, the gamma-corrected illumination is added back into the reflection to achieve color enhancement. Then, the detail enhancement is achieved by processing the reflection components through Laplace pyramids. Finally, the detail-enhanced image and the color-corrected image are reconstructed by weighted fusion to produce the enhanced output image. The subjective and objective performance evaluations show that the proposed method in the paper can enhance the details and global contrast of the dark area images more effectively compared to the contrast algorithm, making the output image with better visual effects. Therefore, the method is an effective method for low light image enhancement and has certain engineering application value.