Remote sensing image change detection (CD) aims to identify changes occurring in bitemporal remote sensing image of the same area at different times. Due to the presence of targets with various scales in the CD scenes, existing methods have designed multiscale encoders to extract change information at different scales. However, these methods do not directly enhance or enrich high-frequency features, resulting in persistent blurry change boundaries. To address this issue, we propose a multiscale enhanced CD method based on detail supplementation, dubbed MEDS-CD. It designs a multiscale edge-guided adaptive filter to extract high-frequency information and compensate for multiscale features. Specifically, a multiscale transformer encoder is employed to extract multiscale features. To capture high-frequency details lost during the encoding process, a multiscale edge-guided adaptive filter is introduced to extract high-frequency details. Furthermore, to fully exploit the restored high-frequency information for change extraction and edge enhancement, we propose a multiscale convolution modulation module and a fuzzy weighting strategy. Finally, the above process is supervised by a hybrid loss function. The proposed method MEDS-CD is validated on three public datasets, LEVIR-CD, WHU-CD, and SYSU-CD. The experimental results show that MEDS-CD achieves better segmentation performance.

Remote sensing spatio-temporal fusion (STF) aims to integrate remote sensing images with complementary temporal and spatial resolutions to enhance image quality with high spatio-temporal resolutions. Conditional diffusion models have shown promising performance in STF by learning data probability distribution to achieve accurate fusion results. However, existing conditional diffusion models for STF directly use the limited image exemplars as conditional signals, limiting their ability to extract coherent features across images with extended temporal gaps and significant resolution differences. To address this challenge, this paper presents a spatio-temporal Dual Priors guided Diffusion framework for STF, named DPDiff. The DPDiff has three key modules including a Prior Extraction Network (PENet), a residual diffusion framework (RDF) and a Difference-Aware Module (DAM). First, a lightweight CNN-based PENet is designed to learn the spatio-temporal dual priors as conditional signals to control the diffusion process. The dual priors include a spatial mapping prior (SMP) between the coarse and fine images at the same time, and a temporal change prior (TCP) between the coarse and fine images at different times. The SMP and the TCP are reciprocal to ensure precise control of the diffusion process. Then, an RDF is adapted to learn distribution differences between high temporal resolution and high spatial resolution images, rather than directly reconstructing the entire image. Third, a DAM is introduced to align the CNN-predicted space with the ground truth space. Experimental results on four benchmark datasets — CIA, LGC, WaterBank, and the global E-SMILE — demonstrate that DPDiff achieves superior generation fidelity and quantitative performance across diverse regions and platforms, outperforming state-of-the-art methods. The source code is available at https://github.com/paul623/DPDiff.

Remote sensing spatiotemporal fusion (STF) aims to integrate remote sensing images with complementary temporal and spatial resolutions to enhance image quality with high spatiotemporal resolutions. Conditional diffusion models have shown promising performance in STF by learning the data probability distribution to achieve accurate fusion results. However, existing conditional diffusion models for STF directly use the limited image exemplars as conditional signals, limiting their ability to extract coherent features across images with extended temporal gaps and significant resolution differences. To address this challenge, this article presents a spatiotemporal dual priors guided diffusion (DPDiff) framework for STF. The DPDiff has three key modules including a prior extraction network (PENet), a residual diffusion framework (RDF), and a difference-aware module (DAM). First, a lightweight convolutional neural network (CNN)-based PENet is designed to learn the spatiotemporal dual priors as conditional signals to control the diffusion process. The dual priors include a spatial mapping prior (SMP) between the coarse and fine images at the same time, and a temporal change prior (TCP) between the coarse and fine images at different times. The SMP and the TCP are reciprocal to ensure precise control of the diffusion process. Then, an RDF is adapted to learn distribution differences between high temporal resolution and high spatial resolution images, rather than directly reconstructing the entire image. Third, a DAM is introduced to align the CNN-predicted space with the ground-truth (GT) space. The experimental results on four benchmark datasets-CIA, LGC, WaterBank, and the global E-SMILE-demonstrate that DPDiff achieves superior generation fidelity and quantitative performance across diverse regions and platforms, outperforming state-of-the-art methods. The source code is available at https://github.com/paul623/DPDiff

Remote sensing spatiotemporal fusion (STF) aims at fusing high temporal-resolution images with high spatial-resolution images to obtain both high temporal and spatial resolution images. Temporal uncertainty remains a critical challenge, as existing methods rely on implicit temporal modeling and thus struggle to handle land-cover or phenological changes over long intervals, leading to texture errors or overly smooth transitions. To overcome this challenge, this paper presents a novel land-cover change inpainting network (LCCINet) for STF. The LCCINet consists of two core stages: the enhancing feature fusion stage and the feature reconstruction stage, which are designed to reconstruct the unchanged and changed regions, respectively. First, to distinguish unchanged and changed regions, we design a change detection module to compute the difference between the prediction and the reference time, generating a change mask. In the fusion stage, we propose a feature mask modulator (FMM) that leverages the mask to modulate the features of the high spatial-resolution image. Then the feature fusion encoder employs multiscale convolutions and spatiotemporal attention to focus the model on fusing features from unchanged areas. Finally, in the reconstruction stage, the image coarse inpainting module (ICM) and image fine inpainting module (IFM) use the mask and rich contextual cues to reconstruct the features of changed regions from coarse to fine. Extensive experiments on three STF benchmarks (CIA, LGC, and E-SMILE) demonstrate that LCCINet consistently outperforms existing approaches, reducing mean absolute error, mean absolute error, and structural similarity assessment measure by 4-11%, and increasing SSIM by 0.003-0.008 over the strongest competing methods. These results confirm the effectiveness of the proposed method in preserving spatial textures and spectral values.

Existing leading Co-saliency Detection (CoD) framework aims to segment the co-salient objects by learning the consensus visual representation of the foreground objects. However, despite different categories, some distractors may have similar appearance to the co-salient objects, such as Apples vs. Bananas have similar color and textures. This makes it challenging to distinguish the distractors only through learning the co-salient object appearance representations. To address this issue, we propose a joint appearance and shape co-representation learner for CoD, dubbed as ASCoD. The ASCoD is composed of a Co-Appearance learning Module (CoAM) and a Co-Shape learning Module (CoSM). The CoAM first learns a co-salient object appearance embedding that encodes the global cross-image and spatial context information. Then, this embedding is set as a co-appearance prototype, which guides the model to enhance the features to highlight the co-salient object regions. Afterwards, we design the CoSM that is a cross-attention module, among which the key and the value encode the shape information from a set of salient tokens dynamically selected by a Co-Shape Prototype generation Module (CSPM). Finally, through jointly optimizing the cascaded CoAM and CoSM, the optimal appearance and shape co-representations are achieved, marrying the merits of both appearance and shape co-representations that are not only robust to co-salient objects appearance variations, but also can well discriminate the co-salient objects from the distractors with similar appearance. Extensive evaluations on three challenging benchmarks including CoCA, CoSOD3k and CoSal2015, demonstrate superiority of the ASCoD to a variety of state-of-the-art CoD methods.

Remote sensing change detection aims to identify changes on the Earth's surface from remote sensing images acquired at different times. However, the identification of changed areas is often hindered by pseudochanges in similar objects, leading to inaccurate identification of change boundaries. To address this issue, we propose a novel network named boundary-guided semantic context network (BSCNet), which decouples features to improve the feature representation ability for changing objects. Specifically, we design a selective context fusion module that selectively fuses semantically rich features by computing the similarity between features from adjacent stages of the backbone network, thereby preventing detailed features from being overwhelmed by contextual information. In addition, to enhance the ability to perceive changes, we design a context fast aggregation module that leverages a pyramid structure to help the model simultaneously extract and fuse detailed and semantic information at different scales, enabling more accurate change detection. Finally, we design a boundary-guided feature fusion module to aggregate edge-level, texture-level, and semantic-level information, which enables the network to represent change regions more comprehensively and precisely. Experimental results on the WHU-CD, LEVIR-CD, and SYSU-CD datasets show that BSCNet achieves F1 scores of 94.92%, 92.19%, and 82.55%, respectively.

Remote sensing spatiotemporal fusion (STF) remains a formidable challenge in scenarios with complex land-cover changes. Existing methods generally use masking mechanisms to cover changed regions, aiming to mitigate alignment biases induced by change information. However, such indiscriminate masking strategies ignore transferable structural cues embedded in pseudo changes, leading to texture collapse and boundary degradation. To address these issues, we propose a change-aware (CA) remote sensing spatiotemporal fusion method via frequency decoupling and symmetric gating, termed CA-STF. Specifically, a wavelet feature modulation module (WFMM) is first designed to perform frequency decoupling, which extracts reusable structural cues from pseudo changes under low-frequency guidance while suppressing spectral perturbations. Subsequently, a spatiotemporal synergistic enhancement module (SSEM) employs a symmetric gating mechanism to selectively fuse the cue-enhanced features, adaptively decoupling changed and unchanged regions. Finally, a mixture-of-experts based spatial restoration module (SRM) decodes the gated features to reconstruct fine-grained textures in changed areas. Experimental results on the public LGC and CIA datasets demonstrate that our CA-STF achieves superior fusion accuracy and visual quality in complex spatiotemporal dynamics.

ABSTRACT Current co‐salient object detection (CoSOD) methods leverage common visual representations to identify recurring foreground objects across image groups. A significant challenge arises when distractors from different categories exhibit high visual similarity to the target objects—such as apples and bananas sharing comparable color and texture—making pure appearance‐based matching prone to failure. To address this issue, we introduce En‐ASCoD, an enhanced architecture that integrates both appearance and shape cues to form a more discriminative consensus representation. The model incorporates a Global Co‐appearance Module (GoAM) to extract group‐level appearance prototypes, a Local Co‐appearance Module (LoAM) that refines these features through contrastive learning within local contexts, and a Co‐shape Module (CoSM) that introduces structural constraints via cross‐attention operating on salient tokens selected through global average pooling. By jointly optimizing these modules in a cascaded manner, En‐ASCoD effectively suppresses visually similar distractors while maintaining robustness to intra‐group appearance variations. Extensive experiments on three challenging benchmarks including CoCA, CoSOD3k, and CoSal2015 show that our method consistently outperforms state‐of‐the‐art alternatives, achieving notable gains in both detection accuracy and generalization.

Existing deep learning-based remote sensing spatiotemporal fusion (STF) relies on a data-driven paradigm without considering the degradation prior modeling from the coarseto fine-resolution images. This makes the learned model easy to overfit to the training dataset, resulting in poor domain generalization across different datasets. To this end, this paper presents a deep frequency degradation prior to STF, dubbed as DeepFDP. The DeepFDP is based on a statistical observation that the frequency feature distributions of the tokens from the coarse- and fine-resolution images in different datasets have a low intra-resolution variance and a high inter-resolution variance with compact well-separated clusters. Therefore, the DeepFDP first designs a frequency proximity module to learn the mapping function between the token frequency representations of the coarse- and fine-resolution images, which is to narrow their feature distribution difference. Since the statistical properties of the token frequency representations are independent of the land-cover classes, the DeepFDP can be learned on a training set of limited image pairs without extra-supervision signals, which has a favorable zero-shot generalization capability across different datasets. Then, to faithfully recover the land-surface details, a high-frequency feature modulation module is designed that uses the fine-resolution image as guidance to progressively learn the multi-scale residual features in a coarse-to-fine fashion, yielding the fused features with rich high-frequency details. Finally, the progressively-fused features at each stage are fed into a hybrid fusion module, yielding the fine-resolution image prediction. Extensive evaluations on LGC and CIA datasets demonstrate favorable performance of the DeepFDP over state-of-the-art methods. Especially, the DeepFDP also shows a good zero-shot generalization performance when training on LGC and testing on CIA.

Remote sensing spatio-temporal fusion (STF) aims at fusing temporally-dense coarse-resolution images and temporally-sparse fine-resolution images to reconstruct high spatio-temporal resolution images. Multi-band remote sensing images are often accepted as inputs for STF that have complementary characteristics for high-fidelity land surface reconstruction, yet the existing STF framework often treats different bands uniformly without considering the statistical correlation between different bands, resulting in unsatisfying results. To address this problem, this paper presents a group-wise semantic-enhanced interactive network for STF, dubbed as GSINet. Based on statistical observations, the feature correlation between the visible-light group and the invisible-light group is weak, while the intra-group correlation is strong. Therefore, the GSINet first separates the inputs into visible-light and invisible-light groups, which are fed into different branches with independent encoders for feature extraction and fusion. Afterwards, to address the issue of land cover changes between the prediction coarse- and reference fine-resolution images, a Semantic-Enhancement Fusion Module (SEFM) is designed to interact with the features from the same group with enhanced semantic information captured in an unsupervised learning manner. Then, the semantic-enhanced fused features from different bands are fed into an Interleaved Cross-attention Module (ICM) for further fusion. Finally, the output fusion features fully encode the intra- and inter-group information, which are fed into the decoder, reconstructing the spatio-temporal high-resolution images. Extensive experiments on CIA and LGC benchmark datasets demonstrate that the GSINet outperforms a variety of state-of-the-art methods in terms of multiple metrics.

Given a coarse-resolution remote sensing image on a prediction date as input, existing spatio-temporal fusion methods commonly use a pair of coarse and fine resolution images that are acquired close to the prediction date. These images serve as references to predict the corresponding fine-resolution image. Recently, this paradigm has been shifted to be unpaired-reference based that only needs one flexible high-resolution image without date restriction as reference. Despite the flexibility of the reference, current work with this paradigm suffers from the following issue: due to the relatively long revisit period, drastic changes caused by cloud pollution or floods may occur in different periods. It results in large land cover changes of losing texture details and even semantic category shift (e.g., from the land to the water), which makes it difficult to obtain enough high-quality reference data, leading to severe model degradation. To solve the above problems, we propose the deformable global-local feature alignment network (DGFANet) for unpaired spatio-temporal fusion, which combines convolutional neural network and transformer to enhance texture and semantic details through global-local alignment. We design a feature alignment module to link the changed region with the surrounding stable region to obtain the global context information. Next, we perform feature fusion using the cross-communication mixture of experts module, which adaptively retains both local features and global representations. Finally, the color consistency loss is proposed to recover the color change of the fused image. In the experimental portion, DGFANet performs equally or demonstrates superior performance compared to the existing state-of-the-art methods in two widely recognized public datasets Colebly irrigated area and lower Gwydir catchment, which has been shown to do so in experimental studies.

Remote sensing change detection (RSCD) aims to identify change areas between bi-temporal images of the same location captured at different points in time. The existing siamese framework adopts a shared-weight strategy to process each bitemporal image independently. Due to the lack of inter-image information interaction, this strategy exhibits limited capability to target change perception and discrimination when faced with small targets or ambiguous changes such as low-covering change areas and irregular morphology in real-world complex scenes. To this end, we propose a one-stream framework using the State Space (SS) Model Mamba to jointly perform feature learning and mixing for RSCD, dubbed as SSCD. Specifically, by leveraging the long-range modeling capability and linear computational complexity of the SS model Mamba, the SSCD leverages a unified approach to feature extraction and information integration through simultaneously processing the bi-temporal images. This enables the model to intensively mutually guide to extract discriminative change features. In addition, to recover more spatial details, we design a texture enhancement module that makes full use of the selective scan modeling capability of the Mamba to enhance the texture features in different directions. Without bells and whistles, our SSCD achieves the state-of-the-art performance on three benchmark datasets including SYSU-CD, LEVIR-CD, and LEVIR+-CD.

Existing leading unsupervised video object segmentation (UVOS) paradigm often leverages a dual-stream architecture with motion and appearance branches, where only the motion cues from optical flow are used as a guide to locating the primary foreground objects. When suffering from challenging factors such as static scenes, fast camera shaking, severe motion blur, etc., the estimated optical flow is noisy with low quality, leading to erroneous primary foreground objects estimation. To address this issue, we propose an open-vocabulary saliency-guided progressive refinement network for UVOS, dubbed as OVSNet. It is observed that most of the primary foreground objects also demonstrate saliency characteristics in the appearance branch. Based on this, our OVSNet complements motion cues with saliency cues predicted by a series of foundation models equipped with favorable zero-shot generalization capabilities. Specifically, we first leverage the off-the-shelf contrastive vision-language pre-training (CLIP) and CLIPSeg to generate an OVS attention map as saliency cues. Then, the saliency cues together with motion cues prompt the segment anything model (SAM) to generate a location map. In the location process, we design two lightweight adapters to fine-tune SAM, which makes SAM well adapt to the downstream UVOS task. Finally, the location map generated by SAM is used to progressively guide object representation refinement in the appearance branch, ultimately achieving an accurate segmentation mask prediction. Extensive evaluations on DAVIS-16, FBMS, and YouTube-Objects demonstrate the favorable performance of our OVSNet over the state-of-the-art methods.

Multispectral image (MSI) and hyperspectral image (HSI) fusion (MHIF) aims to address the challenge of acquiring high-resolution (HR) HSI images. This field combines a low-resolution (LR) HSI with an HR-MSI to reconstruct HR-HSIs. Existing methods directly utilize transformers to perform feature extraction and fusion. Despite the demonstrated success, there exist two limitations: (1) Employing the entire transformer model for feature extraction and fusion fails to fully harness the potential of the transformer in integrating the spectral information of the HSI and spatial information of the MSI. (2) HSIs have a strong spectral correlation and exhibit sparsity in the spatial domain. Existing transformer-based models do not optimize this physical property, which makes their methods prone to spectral distortion. To accomplish these issues, this paper introduces a novel framework for MHIF called a Sparse Mix-Attention Transformer (SMAformer). Specifically, to fully harness the advantages of the transformer architecture, we propose a Spectral Mix-Attention Block (SMAB), which concatenates the keys and values extracted from LR-HSIs and HR-MSIs to create a new multihead attention module. This design facilitates the extraction of detailed long-range information across spatial and spectral dimensions. Additionally, to address the spatial sparsity inherent in HSIs, we incorporated a sparse mechanism within the core of the SMAB called the Sparse Spectral Mix-Attention Block (SSMAB). In the SSMAB, we compute attention maps from queries and keys and select the K highly correlated values as the sparse-attention map. This approach enables us to achieve a sparse representation of spatial information while eliminating spatially disruptive noise. Extensive experiments conducted on three synthetic benchmark datasets, namely CAVE, Harvard, and Pavia Center, demonstrate that the SMAformer method outperforms state-of-the-art methods.

Current mainstream unsupervised video object segmentation (UVOS) approaches typically incorporate optical flow as motion information to locate the primary objects in coherent video frames. However, they fuse appearance and motion information without evaluating the quality of the optical flow. When poor-quality optical flow is used for the interaction with the appearance information, it introduces significant noise and leads to a decline in overall performance. To alleviate this issue, we first employ a quality evaluation module (QEM) to evaluate the optical flow. Then, we select high-quality optical flow as motion cues to fuse with the appearance information, which can prevent poor-quality optical flow from diverting the network’s attention. Moreover, we design an appearance-guided fusion module (AGFM) to better integrate appearance and motion information. Extensive experiments on several widely utilized datasets, including DAVIS-16, FBMS-59, and YouTube-Objects, demonstrate that the proposed method outperforms existing methods.

Previous pure vision paradigm for co-saliency detection (COD) predominantly employs supervised training. The supervisory signals often consist of binary masks or a combination of masks and category labels. However, constrained by limited training samples, these models often suffer from overfitting issue, struggling to generalize to unseen samples. To this end, this paper presents the constrative language-image pretraining-COD (CLIP-COD), a novel language-guided semantic alignment paradigm for COD. The primary objective is to leverage CLIP for aligning concepts between language and images, where the alignment can effectively leverage the powerful language understanding capability of CLIP and transfer its knowledge to image domain, thereby enhancing the model’s zero-shot generalization ability for COD. Firstly, we propose a semantic alignment branch (SAB) that can learn rich knowledge for comprehending images globally. Meanwhile, the SAB can narrow the gap in high-dimensional feature space between the language and image features, transferring the powerful semantic knowledge from CLIP to our model. Subsequently, we devise an intra-group multi-fusion module (IMM) to capture features that integrate group knowledge as dense prompts, providing spatial localization information for subsequent fine segmentation. Finally, we input sparse language prompts and dense mask cues into the pre-trained SAM decoder to obtain the final COD results. Additionally, we further design a transfer optimization adaptor, which can reduce the model training scale, saving computing resource and cost greatly. Extensive experiments on three benchmark datasets, including CoSal2015, CoCA, and CoSOD3k, demonstrate the superior performance of our CLIP-COD to a variety of state-of-the-art methods.

The performance of existing unsupervised video object segmentation methods typically suffers from severe performance degradation on test videos when tested in out-of-distribution scenarios. The primary reason is that the test data in real- world may not follow the independent and identically distribution (i.i.d.) assumption, leading to domain shift. In this paper, we propose a generalizable fourier augmentation method during training to improve the generalization ability of the model. To achieve this, we perform Fast Fourier Transform (FFT) over the intermediate spatial domain features in each layer to yield corresponding frequency representations, including amplitude components (encoding scene-aware styles such as texture, color, contrast of the scene) and phase components (encoding rich semantics). We produce a variety of style features via Gaussian sampling to augment the training data, thereby improving the generalization capability of the model. To further improve the cross-domain generalization performance of the model, we design a phase feature update strategy via exponential moving average using phase features from past frames in an online update manner, which could help the model to learn cross-domain-invariant features. Extensive experiments show that our proposed method achieves the state-of-the-art performance on popular benchmarks.

Existing deep learning based remote sensing change detection (RSCD) methods only rely on binary ground-truth to guide the network learning while neglecting the useful semantic guidance. As a result, the network can be readily misled by irrelevant category changes, leading to degraded performance and slow convergence of the model. To this end, we propose a novel segment anything model (SAM) guided framework, termed as SAM-CD, which mines the rich semantic knowledge from the SAM for RSCD. Specifically, we first employ a transformer encoder to extract multi-scale global features from the bi-temporal images. Meanwhile, we obtain semantic prior masks from the bi-temporal images by providing the SAM with category-relevant text prompts. Then, using the semantic prior masks as constraints, we design a masked attention module (MAM) that generates local features related to the interested categories. Finally, the local and global features are fused and fed into a multi-layer perception (MLP) decoder to obtain the change map. The whole network is trained in an end-to-end manner that can readily encode the rich semantic knowledge of the changed targets to predict an accurate change map. Extensive experiments demonstrate that the proposed SAM-CD achieves state-of-the-art performance on a variety of benchmark datasets.

Existing learning-based remote sensing change detection (RSCD) commonly uses semantic-agnostic binary masks as supervision, which hinders their ability to distinguish between different semantic types of changes, resulting in a noisy change mask prediction. To address this issue, this paper presents a Language-guided semantic clustering framework that can effectively transfer the rich semantic information from the contrastive language-image pretraining (CLIP) model for RSCD, dubbed LSC-CD. The LSC-CD considers the strong zero-shot generalization of the CLIP, which makes it easy to transfer the semantic knowledge from the CLIP into the CD model under semantic-agnostic binary mask supervision. Specifically, the LSC-CD first constructs a category text-prior memory bank based on the dataset statistics and then leverages the CLIP to transform the text in the memory bank into the corresponding semantic embeddings. Afterward, a CLIP adapter module (CAM) is designed to fine-tune the semantic embeddings to align with the change region embeddings from the input bi-temporal images. Next, a semantic clustering module (SCM) is designed to cluster the change region embeddings around the semantic embeddings, yielding the compact change embeddings that are robust to noisy backgrounds. Finally, a lightweight decoder is designed to decode the compact change embeddings, yielding an accurate change mask prediction. Experimental results on three public benchmarks including LEVIR-CD, WHU-CD, and SYSU-CD demonstrate that the proposed LSC-CD achieves state-of-the-art performance in terms of all evaluated metrics.

The goal of low-light image enhancement (LLIE) is to enhance perception to restore normal-light images. The primary emphasis of earlier LLIE methods was on enhancing the illumination while paying less attention to the color distortions and noise in the dark. In comparison to the ground truth, the restored images frequently exhibit inconsistent color and residual noise. To this end, this paper introduces a Wasserstein contrastive regularization method (WCR) for LLIE. The WCR regularizes the color histogram (CH) representation of the restored image to keep its color consistency while removing noise. Specifically, the WCR contains two novel designs including a differentiable CH module (DCHM) and a WCR loss. The DCHM serves as a modular component that can be easily integrated into the network to enable end-to-end learning of the image CH. Afterwards, to ensure color consistency, we utilize the Wasserstein distance (WD) to quantify the resemblance of the learnable CHs between the restored image and the normal-light image. Then, the regularized WD is used to construct the WCR loss, which is a triplet loss and takes the normal-light images as positive samples, the low-light images as negative samples, and the restored images as anchor samples. The WCR loss pulls the anchor samples closer to the positive samples and simultaneously pushes them away from the negative samples so as to help the anchors remove the noise in the dark. Notably, the proposed WCR method was only used for training, and was shown to achieve high performance and high speed inference using lightweight networks. Therefore, it is valuable for real-time applications such as night automatic driving and night reversing image enhancement. Extensive evaluations on benchmark datasets such as LOL, FiveK, and UIEB showed that the proposed WCR method achieves superior performance, outperforming existing state-of-the-art methods.