摘要
在模糊光环境下,由于光线条件不佳,亮度不均匀等因素的影响,图像中色彩偏移和失真增大了特征增强难度。为了改善模糊光图像的视觉传达效果,提出了一种模糊光图像特征融合生成对抗网络增强方法。将模糊光图转换至YCbCr颜色空间,并以Y分量为参考对图像实施白平衡调节,降低后续模糊光图像增强复杂度。以调节后的图像为生成对抗网络的输入,通过生成器内色彩注意力机制的添加,生成关注图像色彩信息和细节信息的特征图;利用AFF模块融合生成的特征图后,通过设计损失函数,进行判断器与生成器对抗性训练,通过不断训练优化图像真实性和整体质量,实现模糊光图像增强。结果表明,所提方法能够精准调节模糊光图像的白平衡。在保证图像细节信息的前提下,所提方法实现了模糊光图像质量、色彩、细节的整体增强,为图像信息的有效利用提供技术支持。
In the fuzzy light environment,due to poor light conditions,uneven brightness and other factors,color deviation and distortion in the image increase the difficulty of feature enhancement.In order to improve the visual communication effect of fuzzy light image,a method of fuzzy light image feature fusion to generate confrontation net-work enhancement is proposed.Convert the blurred light image to YCbCr color space,and use the Y component as a reference to adjust the white balance of the image to reduce the complexity of subsequent blurred light image enhance-ment.With the adjusted image as the input of the generated confrontation network,a feature map focusing on the color information and details of the image is generated by adding the color attention mechanism in the generator;After using the feature map generated by the fusion of the AFF module,through the design of the loss function,the adversary training between the judge and the generator is carried out.Through continuous training,the authenticity and overall quality of the image are optimized to achieve the enhancement of the blurred light image.The results show that the proposed method can accurately adjust the white balance of the blurred light image.On the premise of ensuring the image detail information,the proposed method realizes the overall enhancement of fuzzy light image quality,color and detail,and provides technical support for the effective use of image information.
作者
侯梦
乔迁
HOU Meng;QIAO Qian(Modern College,Northwest University,Shaanxi Xi'an 710076,China;Xijing University,Shaanxi Xi'an 710123,China)
出处
《计算机仿真》
2025年第9期492-496,共5页
Computer Simulation
关键词
颜色空间
白平衡调节
注意力机制
特征图融合
模糊光图像
Color space
White balance adjustment
Attention mechanism
Feature map fusion
Blurred light images
