一种基于YCbCr色彩空间的无监督学习弱光图像增强器OA

In recent years,deep learning methods have been widely used in the field of low-light image enhancement and have achieved good results.The main purpose of low-light image enhancement is to restore the visibility of the object of interest.However,traditional methods often overlook issues such as noise and artifacts,resulting in a significant degradation of the quality of the enhanced image.In this paper,an unsupervised learning low-light image enhancer based on YCbCr colour space was proposed.The network divided the low-light image enhancement into two parts.First,a brightness adjustment network(BAN)was designed and the luminance enhancement residual(LER)block was used to adjust the luminance.In addition,the CbCr channels were enhanced through a color recovery network(CRN),which captured detailed information from the chrominance channels and integrated both global and local feature information.Finally,the enhanced luminance and chrominance images were spliced in the channel direction and converted to RGB images through colour space.Experimental results show that the proposed method can better handle various types of low-light images and outperforms most of the existing low-light image enhancement methods.The method proposed in this paper has been extensively experimented on four publicly available datasets and has shown good performance compared with other state-of-the-art competitors in terms of reference/non-reference quality assessment.