Article
A COMPARATIVE STUDY ON CNN-BASED LOW-LIGHT IMAGE ENHANCEMENT
Low- light image improvement is a grueling task that has attracted considerable attention. filmland taken in low- light conditions frequently have bad visual quality. To address the problem, we regard the low- light improvement as a residual literacy problem that's to estimate the residual between low- and normal- light images. In this paper, we propose a new Deep Lightening Network (DLN) that benefits from the recent development of Convolutional Neural Networks (CNNs). The proposed DLN consists of several Lightening Back protuberance (LBP) blocks. The LBPs perform lightening and darkening processes iteratively to learn the residual for normal- light estimations. To effectively use the original and global features, we also propose a point Aggregation (FA) block that adaptively fuses the results of different LBPs. We estimate the proposed system on different datasets. Numerical results show that our proposed DLN approach outperforms other styles under both objective and private criteria
Full Text Attachment
