摘要
低照条件下拍摄的图像质量退化、能见度差和边缘模糊,从而严重影响图像在目标检测、识别、分类等高级视觉任务中的表现.针对这种问题提出一种融合同态滤波与注意力机制的RetinexNet图像增强算法.该算法先行将图像转换至HSV颜色空间,采用融合同态滤波与注意力机制的改进RetinexNet策略,对亮度分量V进行调节和对比度增强处理.随后,基于自适应γ校正模式,结合增强后的亮度强度对饱和度分量S进行色彩平衡调节.最后,将处理后的图像转换回RGB空间,得到增强后的图像.实验结果表明,该方法在显著提升图像亮度和对比度的同时,能增强图像细节和轮廓表现,保持较好的色彩自然度,改善视觉效果.
Images captured under low-light conditions suffer from quality degradation,poor visibility,and edge blurring,which seriously affect the performance of images in advanced visual tasks such as target detection,recognition,and classification.A RetinexNet image enhancement algorithm that combines homomorphic filtering and attention mechanism is proposed to address these problems.The algorithm first converts the image to HSV color space,and adopts an improved RetinexNet strategy that integrates homomorphic filtering and attentional mechanism to adjust the luminance component V and enhance the contrast.Subsequently,based on the adaptive gamma correction mode,the saturation component S is adjusted to the color balance by combining the enhanced luminance intensity.Finally,the processed image is converted back to RGB space to obtain the enhanced low-light image.The experimental results show that this method can enhance the image details and contour performance,maintain a better color naturalness and improve the visual effect while significantly improving the image brightness and contrast.
作者
刘雨轩
罗海月
蒋滔
刘呈滢
孙艺
廖雪花
LIU Yuxuan;LUO Haiyue;JIANG Tao;LIU Chengying;SUN Yi;LIAO Xuehua(College of Computer Science,Sichuan Normal University,Chengdu 610101,Sichuan;College of Physics and Electronic Engineering,Sichuan Normal University,Chengdu 610101,Sichuan)
出处
《四川师范大学学报(自然科学版)》
2026年第2期237-245,共9页
Journal of Sichuan Normal University(Natural Science)
基金
国家社会科学基金一般项目(20BMZ092)。
