Semantically Contrastive Learning for Low-Light Image Enhancement

Abstract

Low-light image enhancement (LLE) remains challenging due to the unfavorable prevailing low-contrast and weak-visibility problems of single RGB images. In this paper, we respond to the intriguing learning-related question -- if leveraging both accessible unpaired over/underexposed images and high-level semantic guidance, can improve the performance of cutting-edge LLE models? Here, we propose an effective semantically contrastive learning paradigm for LLE (namely SCL-LLE). Beyond the existing LLE wisdom, it casts the image enhancement task as multi-task joint learning, where LLE is converted into three constraints of contrastive learning, semantic brightness consistency, and feature preservation for simultaneously ensuring the exposure, texture, and color consistency. SCL-LLE allows the LLE model to learn from unpaired positives (normal-light)/negatives (over/underexposed), and enables it to interact with the scene semantics to regularize the image enhancement network, yet the interaction of high-level semantic knowledge and the low-level signal prior is seldom investigated in previous methods. Training on readily available open data, extensive experiments demonstrate that our method surpasses the state-of-the-arts LLE models over six independent cross-scenes datasets. Moreover, SCL-LLE's potential to benefit the downstream semantic segmentation under extremely dark conditions is discussed. Source Code: https://github.com/LingLIx/SCL-LLE.

FDMLNet: A Frequency-Division and Multiscale Learning Network for Enhancing Low-Light Image

“…Comparison approaches: We carefully selected 12 state-of-the-art approaches as comparison methods for validating the superiority of this FDMLNet for light enhancement. These selected methods contained three traditional methods, i.e., LR3M [ 18 ], simultaneous reflection and illumination estimation (SRIE) [ 19 ], and the bioinspired multiexposure fusion framework (BIMEF) [ 20 ]; seven supervised-learning-based methods, i.e., RetinexNet [ 54 ], deep stacked Laplacian restorer (DSLR) [ 49 ], KinD [ 28 ], DLN [ 14 ], DRBN [ 59 ], SCL-LLE [ 52 ], and MIRNet [ 65 ]; an unsupervised-learning-based method, i.e., EnlightenGAN [ 29 ]; and a zero-reference-learning-based method, i.e., Zero DCE++ [ 1 ]. Notably, three traditional methods were coded in Matlab and the other eight comparison methods were coded in Python and Pytorch.…”

“…KinD [ 28 ] failed to recover the inherent details and introduced unsatisfactory color casts in local dark regions of the image. SCL-LLE [ 52 ] generated undesired images with an unnatural visual experience (observed in picture g in Figure 10 ). MIRNet [ 52 ] succeeded in improving the image brightness, but the enhanced images exhibited a color deviation and low contrast.…”

“…EEMEFN [ 47 ] and TBEFN [ 48 ] generated normal light pictures by fusing multiexposure images. Subsequently, the pyramid network [ 49 , 50 ], residual network [ 51 ], image semantic network [ 52 ], semantically contrastive learning [ 52 ], and recursive learning network [ 53 ] were introduced to enhance the feature representation and extraction of the previously reported model. Moreover, the Retinex theory and learning-based model were combined to make the proposed methods enjoy an appealing performance.…”

SGINet: Toward Sufficient Interaction Between Single Image Deraining and Semantic Segmentation

“…Few existing methods can deal with the degraded images by low-and high-level vision interaction [15,17,18,23,24]. These methods usually incorporate the high-level tasks into the overall framework, providing more image priors and further benefiting degraded image restoration.…”

“…These methods usually incorporate the high-level tasks into the overall framework, providing more image priors and further benefiting degraded image restoration. For example, SCL-LLE [17] proposes an effective semantically contrastive learning framework for low-light image enhancement, which embeds the semantic segmentation task into the enhancement model. As a result, the offered high-level ... semantic knowledge can be used to guide the illumination enhancement process.…”