高级检索
张杰, 周浦城, 薛模根. 基于方向性全变分Retinex的低照度图像增强[J]. 计算机辅助设计与图形学学报, 2018, 30(10): 1943-1953. DOI: 10.3724/SP.J.1089.2018.16965
引用本文: 张杰, 周浦城, 薛模根. 基于方向性全变分Retinex的低照度图像增强[J]. 计算机辅助设计与图形学学报, 2018, 30(10): 1943-1953. DOI: 10.3724/SP.J.1089.2018.16965
Zhang Jie, Zhou Pucheng, Xue Mogen. Low-light Image Enhancement Based on Directional Total Variation Retinex[J]. Journal of Computer-Aided Design & Computer Graphics, 2018, 30(10): 1943-1953. DOI: 10.3724/SP.J.1089.2018.16965
Citation: Zhang Jie, Zhou Pucheng, Xue Mogen. Low-light Image Enhancement Based on Directional Total Variation Retinex[J]. Journal of Computer-Aided Design & Computer Graphics, 2018, 30(10): 1943-1953. DOI: 10.3724/SP.J.1089.2018.16965

基于方向性全变分Retinex的低照度图像增强

Low-light Image Enhancement Based on Directional Total Variation Retinex

  • 摘要: 为了提高低照度图像的视觉质量,提出一种基于方向性全变分Retinex的低照度图像增强方法.首先利用基于灰像素的色彩恒常方法恢复低照度图像的真实场景颜色;然后构造方向性相对全变分约束,据此估计照度图像;接着在计算反射图像时,利用方向性梯度残余最小约束来遏制增强过程中放大的噪声和伪影;最后根据当前像素亮度与其邻域平均亮度的关系拉伸局部亮度,得到最终的增强结果.采用从互联网上搜集的多幅低照度图像进行实验,结果表明,与一些当前国际先进水平的方法相比,文中方法得到的图像质量更佳、视觉效果更好.

     

    Abstract: In order to improve the visual quality of low-light image,a kind of low-light image enhancement method based on directional total variation Retinex was proposed.Firstly,the true scene color of low-light image was restored by grey pixels-based color constancy pretreatment.And then,the directional relative total variation constraint was constructed,so as to estimate the illumination map.In order to suppress the amplification of noise and artifact during the enhancement procedure,the directional gradient residual minimization constraint was used to calculate the reflection map.Finally,considering the relationship between the current pixel’s brightness and its neighboring pixels’average brightness,the final enhancement image was obtained by stretching the local brightness.Numerous experiments on various low-light images collected from Internet demonstrate that the proposed method can obtain more pleasing results compared with the state-of-the-art methods.

     

/

返回文章
返回