摘要
在低照度条件下拍摄的图像具有对比度低,亮度低,细节缺失等质量缺陷,给图像处理带来困难;提出一种改进零参考深度曲线低照度图像增强算法,通过在空间一致性损失函数中引入与卷积核大小相关参数,统一了不同尺寸图像的增强效果;将颜色不变损失、照明平滑损失函数与输入图像类型关联,使其增强效果的峰值信噪比提高17.75%,对比度提高26.75%;通过使用对称式卷积结构,解决原算法计算量大的问题;通过使用MobileNetV2轻量化网络对零参考深度网络(Zero-DCE)进行了优化,减少网络模型计算复杂度的同时保证模型较好的增强效果。
Low luminous image has some quality defects of low contrast, low brightness and missing details, it brings difficulties to image processing. An improved algorithm for zero reference depth curve and low luminous image enhancement is proposed. By introducing the related parameters of convolution kernel into the spatial consistency loss function, the enhancement effect of different images is unified. The loss functions of color invariant loss and light smoothing are related to the input image type, so that the peak signal-to-noise ratio of enhancement effect is increased by 17.75%,and the contrast by 26.75%. Symmetric convolution structures are used to solve the original algorithm has a large amount of calculation. A MobileNetV2 lightweight network model is used to optimize the zero reference depth network(Zero-DCE), which reduces the computational complexity and ensures the better enhancement effect of the network model.
作者
陈从平
张力
江高勇
凌阳
戴国洪
CHEN Congping;ZHANG Li;JIANG Gaoyong;LING Yang;DAI Guohong(School of Mechanics and Rail Transit,Changzhou University,Changzhou 213164,China;School of Materials Science and Engineering,Changzhou University,Changzhou 213164,China)
出处
《计算机测量与控制》
2023年第1期209-214,221,共7页
Computer Measurement &Control
基金
国家自然科学基金(51475266)
江苏省产业前瞻与关键核心技术-碳达峰碳中和科技创新专项资金项目(BE2022044)。
