首页|基于上下文注意力的自适应低光照图像增强网络

基于上下文注意力的自适应低光照图像增强网络

扫码查看
在照明环境不良的情况下获取到的图像具有可见性低、对比度差、色彩失真以及噪声等问题,严重地干扰了其在后期图像处理中的使用.为了改善这种情况下采集图像的质量,以获得图像中更多有用的信息,针对上述存在的问题,提出了基于上下文注意力的自适应低光照图像增强方法.该方法首先通过卷积对图像进行初步的特征提取,获得四个不同上下文信息的特征;再利用含有像素和通道注意力机制的特征增强模块对提取出的特征进行提炼;最后将学习到的增强特征输入自适应融合调节模块与前面的浅层特征进行融合调节,恢复出符合真实光照的图片.实验结果表明:相比较于一些经典算法,所提出的方法在主观视觉感受上及客观评价指标(PSNR,SSIM,MAE)上均能够得到较好提升.
Adaptive Low-light Image Enhancement Network Based on Contextual Attention
The images collected from low light environment suffer from various challenges including low visibility,poor con-trast,color distortion and noise,which significantly affect subsequent applications in real world.In order to improve the quality of the captured image in this case,a low-light image enhancement method is proposed to relieve the issues,which designs an adaptive context attention mechanism.Specifically,the method firstly performs preliminary feature extraction on images through convolution-al decoder,and obtains the features of four different context information.Then,it uses the feature enhancement module with pix-el-wise and channel-wise attention to refine the extracted features.Finally,the learned enhanced features are input into the adap-tive fusion adjustment module,which integrates the previous shallow features with the refined one.The experimental results show that the proposed method can achieve significant improvement on both metrics of subjective visual perception and objective evalua-tion indicators(PSNR,SSIM,MAE),comparing with recent solutions.

low-light image enhancementcontextual attentionadaptive fusionimage processing

黄金鑫、靳雨桐、宋慧慧

展开 >

南京信息工程大学大气环境与装备技术协同创新中心江苏省大数据分析技术重点实验室 南京 210044

低光照图像增强 上下文注意力 自适应融合 图像处理

2024

计算机与数字工程
中国船舶重工集团公司第七0九研究所

计算机与数字工程

CSTPCD
影响因子:0.355
ISSN:1672-9722
年,卷(期):2024.52(11)
  • 1