细节保留与亮度融合的微光图像增强算法

Low-light Image Enhancement Based on Detail Preservation and Brightness Fusion

  • 摘要: 微光环境下CMOS相机拍摄的图像存在对比度低、噪声大和细节模糊等问题,目前的增强算法虽然可以提升图像亮度,却没有考虑保留原始图像的细节信息,为此,本文提出了一种细节保留的微光图像增强方法。首先,使用自适应滤波器对原始微光图像中的每个通道进行处理,以改善噪声对图像的影响。然后,通过滤波后的图像得到细节层和基本层,其中细节层用于保留图像的纹理,基本层则用于增强图像的亮度。细节层的求取通过粗照度和光照分量得到,而基本层亮度增强则使用Alpha算法融合原图、伽马校正结果和伽马先验校正结果得到。最后,将细节层和基本层的亮度通道进行融合,以获得增强后的图像。实验结果表明,经过细节融合处理后的图像,其平均梯度与信息熵均实现了显著提升。此外,与选取的5种方法相比,SSIM指标均高于其他算法,表明本文方法增强后的图像失真小。同时,在BRISQUE和PSNR指标上均取得了较好的效果。

     

    Abstract: Images captured by complementary metal-oxide semiconductor (CMOS) cameras under low-light environments have problems, including low contrast, high noise, blurred details, and insufficient enhancement algorithms. Although CMOS cameras can improve image brightness, they do not consider how to retain the detailed information of the original image. Therefore, this paper proposes a detail-preserving low-light image enhancement method. First, each channel in the original low-light image was processed using an adaptive filter to improve the effect of noise on the image. Subsequently, a detail layer and a basic layer are obtained from the filtered image. The detail layer is used to preserve the texture of the image, and the basic layer is used to enhance its brightness. Moreover, the detail layer is obtained via coarse illumination and light components, whereas the basic layer brightness enhancement is obtained using the alpha algorithm to fuse the original image, gamma correction result, and gamma priori correction result. Finally, the luminance channels of the detailed and basic layers are fused to obtain an enhanced image. The experimental results showed that the average gradient and information entropy of the fused detailed image were significantly improved. In addition, among the five selected methods, the SSIM index was higher than those of other algorithms, indicating less distortion in the enhanced images. Good results were achieved for both the BRISQUE and PSNR indicators.

     

/

返回文章
返回