Munchurl Kim

Jooyoung Lee, Seyoon Jeong, Munchurl Kim

Recently, many neural network-based image compression methods have shown promising results superior to the existing tool-based conventional codecs. However, most of them are often trained as separate models for different target bit rates, thus increasing the model complexity. Therefore, several studies have been conducted for learned compression that supports variable rates with single models, but they require additional network modules, layers, or inputs that often lead to complexity overhead, or do not provide sufficient coding efficiency. In this paper, we firstly propose a selective compression method that partially encodes the latent representations in a fully generalized manner for deep learning-based variable-rate image compression. The proposed method adaptively determines essential representation elements for compression of different target quality levels. For this, we first generate a 3D importance map as the nature of input content to represent the underlying importance of the representation elements. The 3D importance map is then adjusted for different target quality levels using importance adjustment curves. The adjusted 3D importance map is finally converted into a 3D binary mask to determine the essential representation elements for compression. The proposed method can be easily integrated with the existing compression models with a negligible amount of overhead increase. Our method can also enable continuously variable-rate compression via simple interpolation of the importance adjustment curves among different quality levels. The extensive experimental results show that the proposed method can achieve comparable compression efficiency as those of the separately trained reference compression models and can reduce decoding time owing to the selective compression. The sample codes are publicly available at https://github.com/JooyoungLeeETRI/SCR.

Andrey Ignatov, Radu Timofte, Jin Zhang et al.

As mobile cameras with compact optics are unable to produce a strong bokeh effect, lots of interest is now devoted to deep learning-based solutions for this task. In this Mobile AI challenge, the target was to develop an efficient end-to-end AI-based bokeh effect rendering approach that can run on modern smartphone GPUs using TensorFlow Lite. The participants were provided with a large-scale EBB! bokeh dataset consisting of 5K shallow / wide depth-of-field image pairs captured using the Canon 7D DSLR camera. The runtime of the resulting models was evaluated on the Kirin 9000's Mali GPU that provides excellent acceleration results for the majority of common deep learning ops. A detailed description of all models developed in this challenge is provided in this paper.

Jongmin Park, Jeonghwan Yun, Minh-Quan Viet Bui et al.

Dynamic scene reconstruction from monocular videos remains highly challenging, as existing methods often struggle to balance global structural coherence and local fine-grained details under limited multi-view cues. To address this challenge, we propose WebSpline, a novel dynamic 3D Gaussian framework that enables structurally coherent and high-fidelity reconstruction from monocular videos with fast rendering. The core of WebSpline is the Structure-Informed Spline (SIS) representation, which models each dynamic Gaussian trajectory using a learnable cubic Hermite spline whose motion is structurally organized with an auxiliary Structural Proxy Graph (SPG). The proposed framework is optimized in two stages: (i) in the first stage, the SPG is initialized from 2D point tracks and refined with temporal rigidity regularization to establish structural coherence for moving objects across the sequence; and (ii) in the second stage, the SIS representation is initialized from the refined SPG and optimized under both spatial and structural neighborhood constraints. At inference, Gaussian motion is obtained solely by evaluating the learned SIS, enabling fast rendering. Extensive experiments on the challenging monocular dynamic scene benchmarks, iPhone and NVIDIA, demonstrate that our WebSpline achieves state-of-the-art rendering quality while rendering over 10 times faster than WorldTree, the second-best method on the iPhone dataset.

Agus Gunawan, Soo Ye Kim, Hyeonjun Sim et al.

This paper firstly presents old photo modernization using multiple references by performing stylization and enhancement in a unified manner. In order to modernize old photos, we propose a novel multi-reference-based old photo modernization (MROPM) framework consisting of a network MROPM-Net and a novel synthetic data generation scheme. MROPM-Net stylizes old photos using multiple references via photorealistic style transfer (PST) and further enhances the results to produce modern-looking images. Meanwhile, the synthetic data generation scheme trains the network to effectively utilize multiple references to perform modernization. To evaluate the performance, we propose a new old photos benchmark dataset (CHD) consisting of diverse natural indoor and outdoor scenes. Extensive experiments show that the proposed method outperforms other baselines in performing modernization on real old photos, even though no old photos were used during training. Moreover, our method can appropriately select styles from multiple references for each semantic region in the old photo to further improve the modernization performance.

Soo Ye Kim, Jianming Zhang, Simon Niklaus et al.

Depth maps are used in a wide range of applications from 3D rendering to 2D image effects such as Bokeh. However, those predicted by single image depth estimation (SIDE) models often fail to capture isolated holes in objects and/or have inaccurate boundary regions. Meanwhile, high-quality masks are much easier to obtain, using commercial auto-masking tools or off-the-shelf methods of segmentation and matting or even by manual editing. Hence, in this paper, we formulate a novel problem of mask-guided depth refinement that utilizes a generic mask to refine the depth prediction of SIDE models. Our framework performs layered refinement and inpainting/outpainting, decomposing the depth map into two separate layers signified by the mask and the inverse mask. As datasets with both depth and mask annotations are scarce, we propose a self-supervised learning scheme that uses arbitrary masks and RGB-D datasets. We empirically show that our method is robust to different types of masks and initial depth predictions, accurately refining depth values in inner and outer mask boundary regions. We further analyze our model with an ablation study and demonstrate results on real applications. More information can be found at https://sooyekim.github.io/MaskDepth/ .

Juan Luis Gonzalez Bello, Jaeho Moon, Munchurl Kim

Recently, much attention has been drawn to learning the underlying 3D structures of a scene from monocular videos in a fully self-supervised fashion. One of the most challenging aspects of this task is handling the independently moving objects as they break the rigid-scene assumption. For the first time, we show that pixel positional information can be exploited to learn SVDE (Single View Depth Estimation) from videos. Our proposed moving object (MO) masks, which are induced by shifted positional information (SPI) and referred to as `SPIMO' masks, are very robust and consistently remove the independently moving objects in the scenes, allowing for better learning of SVDE from videos. Additionally, we introduce a new adaptive quantization scheme that assigns the best per-pixel quantization curve for our depth discretization. Finally, we employ existing boosting techniques in a new way to further self-supervise the depth of the moving objects. With these features, our pipeline is robust against moving objects and generalizes well to high-resolution images, even when trained with small patches, yielding state-of-the-art (SOTA) results with almost 8.5x fewer parameters than the previous works that learn from videos. We present extensive experiments on KITTI and CityScapes that show the effectiveness of our method.

Jooyoung Lee, Se Yoon Jeong, Munchurl Kim

Unlike fixed- or variable-rate image coding, progressive image coding (PIC) aims to compress various qualities of images into a single bitstream, increasing the versatility of bitstream utilization and providing high compression efficiency compared to simulcast compression. Research on neural network (NN)-based PIC is in its early stages, mainly focusing on applying varying quantization step sizes to the transformed latent representations in a hierarchical manner. These approaches are designed to compress only the progressively added information as the quality improves, considering that a wider quantization interval for lower-quality compression includes multiple narrower sub-intervals for higher-quality compression. However, the existing methods are based on handcrafted quantization hierarchies, resulting in sub-optimal compression efficiency. In this paper, we propose an NN-based progressive coding method that firstly utilizes learned quantization step sizes via learning for each quantization layer. We also incorporate selective compression with which only the essential representation components are compressed for each quantization layer. We demonstrate that our method achieves significantly higher coding efficiency than the existing approaches with decreased decoding time and reduced model size. The source code is publicly available at https://github.com/JooyoungLeeETRI/DeepHQ

Jeonghyeok Do, Yun Chen, Geunhyuk Youk et al.

The landscape of skeleton-based action representation learning has evolved from Contrastive Learning (CL) to Masked Auto-Encoder (MAE) architectures. However, each paradigm faces inherent limitations: CL often overlooks fine-grained local details, while MAE is burdened by computationally heavy decoders. Moreover, MAE suffers from severe computational asymmetry -- benefiting from efficient masking during pre-training but requiring exhaustive full-sequence processing for downstream tasks. To resolve these bottlenecks, we propose SLiM (Skeleton Less is More), a novel unified framework that harmonizes masked modeling with contrastive learning via a shared encoder. By eschewing the reconstruction decoder, SLiM not only eliminates computational redundancy but also compels the encoder to capture discriminative features directly. SLiM is the first framework with decoder-free masked modeling of representative learning. Crucially, to prevent trivial reconstruction arising from high skeletal-temporal correlation, we introduce semantic tube masking, alongside skeletal-aware augmentations designed to ensure anatomical consistency across diverse temporal granularities. Extensive experiments demonstrate that SLiM consistently achieves state-of-the-art performance across all downstream protocols. Notably, our method delivers this superior accuracy with exceptional efficiency, reducing inference computational cost by 7.89x compared to existing MAE methods.

Jongmin Park, Jooyoung Lee, Munchurl Kim

Recently, neural network (NN)-based image compression studies have actively been made and has shown impressive performance in comparison to traditional methods. However, most of the works have focused on non-scalable image compression (single-layer coding) while spatially scalable image compression has drawn less attention although it has many applications. In this paper, we propose a novel NN-based spatially scalable image compression method, called COMPASS, which supports arbitrary-scale spatial scalability. Our proposed COMPASS has a very flexible structure where the number of layers and their respective scale factors can be arbitrarily determined during inference. To reduce the spatial redundancy between adjacent layers for arbitrary scale factors, our COMPASS adopts an inter-layer arbitrary scale prediction method, called LIFF, based on implicit neural representation. We propose a combined RD loss function to effectively train multiple layers. Experimental results show that our COMPASS achieves BD-rate gain of -58.33% and -47.17% at maximum compared to SHVC and the state-of-the-art NN-based spatially scalable image compression method, respectively, for various combinations of scale factors. Our COMPASS also shows comparable or even better coding efficiency than the single-layer coding for various scale factors.

Geunhyuk Youk, Jihyong Oh, Munchurl Kim

Real-world video restoration is plagued by complex degradations from motion coupled with dynamically varying exposure - a key challenge largely overlooked by prior works and a common artifact of auto-exposure or low-light capture. We present FMA-Net++, a framework for joint video super-resolution and deblurring that explicitly models this coupled effect of motion and dynamically varying exposure. FMA-Net++ adopts a sequence-level architecture built from Hierarchical Refinement with Bidirectional Propagation blocks, enabling parallel, long-range temporal modeling. Within each block, an Exposure Time-aware Modulation layer conditions features on per-frame exposure, which in turn drives an exposure-aware Flow-Guided Dynamic Filtering module to infer motion- and exposure-aware degradation kernels. FMA-Net++ decouples degradation learning from restoration: the former predicts exposure- and motion-aware priors to guide the latter, improving both accuracy and efficiency. To evaluate under realistic capture conditions, we introduce REDS-ME (multi-exposure) and REDS-RE (random-exposure) benchmarks. Trained solely on synthetic data, FMA-Net++ achieves state-of-the-art accuracy and temporal consistency on our new benchmarks and GoPro, outperforming recent methods in both restoration quality and inference speed, and generalizes well to challenging real-world videos.

Jongmin Park, Minh-Quan Viet Bui, Juan Luis Gonzalez Bello et al.

Feed-forward 3D Gaussian Splatting (3DGS) enables efficient one-pass scene reconstruction, providing 3D representations for novel view synthesis without per-scene optimization. However, existing methods typically predict pixel-aligned primitives per-view, producing an excessive number of primitives in dense-view settings and offering no explicit control over the number of predicted Gaussians. To address this, we propose EcoSplat, the first efficiency-controllable feed-forward 3DGS framework that adaptively predicts the 3D representation for any given target primitive count at inference time. EcoSplat adopts a two-stage optimization process. The first stage is Pixel-aligned Gaussian Training (PGT) where our model learns initial primitive prediction. The second stage is Importance-aware Gaussian Finetuning (IGF) stage where our model learns rank primitives and adaptively adjust their parameters based on the target primitive count. Extensive experiments across multiple dense-view settings show that EcoSplat is robust and outperforms state-of-the-art methods under strict primitive-count constraints, making it well-suited for flexible downstream rendering tasks.

Taewoo Suh, Sungpyo Kim, Jongmin Park et al.

Feed-forward 3D Gaussian Splatting (FF-3DGS) emerges as a fast and robust solution for sparse-view 3D reconstruction and novel view synthesis (NVS). However, existing FF-3DGS methods are built on incorrect screen-space dilation filters, causing severe rendering artifacts when rendering at out-of-distribution sampling rates. We firstly propose an FF-3DGS model, called AA-Splat, to enable robust anti-aliased rendering at any resolution. AA-Splat utilizes an opacity-balanced band-limiting (OBBL) design, which combines two components: a 3D band-limiting post-filter integrates multi-view maximal frequency bounds into the feed-forward reconstruction pipeline, effectively band-limiting the resulting 3D scene representations and eliminating degenerate Gaussians; an Opacity Balancing (OB) to seamlessly integrate all pixel-aligned Gaussian primitives into the rendering process, compensating for the increased overlap between expanded Gaussian primitives. AA-Splat demonstrates drastic improvements with average 5.4$\sim$7.5dB PSNR gains on NVS performance over a state-of-the-art (SOTA) baseline, DepthSplat, at all resolutions, between $4\times$ and $1/4\times$. Code will be made available.

Minh-Quan Viet Bui, Jaeho Moon, Munchurl Kim

While 3D Vision Foundation Models (3DVFMs) have demonstrated remarkable zero-shot capabilities in visual geometry estimation, their direct application to generalizable novel view synthesis (NVS) remains challenging. In this paper, we propose AirSplat, a novel training framework that effectively adapts the robust geometric priors of 3DVFMs into high-fidelity, pose-free NVS. Our approach introduces two key technical contributions: (1) Self-Consistent Pose Alignment (SCPA), a training-time feedback loop that ensures pixel-aligned supervision to resolve pose-geometry discrepancy; and (2) Rating-based Opacity Matching (ROM), which leverages the local 3D geometry consistency knowledge from a sparse-view NVS teacher model to filter out degraded primitives. Experimental results on large-scale benchmarks demonstrate that our method significantly outperforms state-of-the-art pose-free NVS approaches in reconstruction quality. Our AirSplat highlights the potential of adapting 3DVFMs to enable simultaneous visual geometry estimation and high-quality view synthesis.

Woojin Lee, Hyugjae Chang, Jaeho Moon et al.

Weakly supervised oriented object detection (WS-OOD) has gained attention as a cost-effective alternative to fully supervised methods, providing both efficiency and high accuracy. Among weakly supervised approaches, horizontal bounding box (HBox)-supervised OOD stands out for its ability to directly leverage existing HBox annotations while achieving the highest accuracy under weak supervision settings. This paper introduces adaptive bounding box scaling and symmetry-prior-based orientation prediction, called ABBSPO, a framework for WS-OOD. Our ABBSPO addresses limitations of previous HBox-supervised OOD methods, which compare ground truth (GT) HBoxes directly with the minimum circumscribed rectangles of predicted RBoxes, often leading to inaccurate scale estimation. To overcome this, we propose: (i) Adaptive Bounding Box Scaling (ABBS), which appropriately scales GT HBoxes to optimize for the size of each predicted RBox, ensuring more accurate scale prediction; and (ii) a Symmetric Prior Angle (SPA) loss that exploits inherent symmetry of aerial objects for self-supervised learning, resolving issues in previous methods where learning collapses when predictions for all three augmented views (original, rotated, and flipped) are consistently incorrect. Extensive experimental results demonstrate that ABBSPO achieves state-of-the-art performance, outperforming existing methods.

Wonyong Seo, Jaeho Moon, Jaehyup Lee et al.

Propagation-based video editing enables precise user control by propagating a single edited frame into following frames while maintaining the original context such as motion and structures. However, training such models requires large-scale, paired (source and edited) video datasets, which are costly and complex to acquire. Hence, we propose the PropFly, a training pipeline for Propagation-based video editing, relying on on-the-Fly supervision from pre-trained video diffusion models (VDMs) instead of requiring off-the-shelf or precomputed paired video editing datasets. Specifically, our PropFly leverages one-step clean latent estimations from intermediate noised latents with varying Classifier-Free Guidance (CFG) scales to synthesize diverse pairs of 'source' (low-CFG) and 'edited' (high-CFG) latents on-the-fly. The source latent serves as structural information of the video, while the edited latent provides the target transformation for learning propagation. Our pipeline enables an additional adapter attached to the pre-trained VDM to learn to propagate edits via Guidance-Modulated Flow Matching (GMFM) loss, which guides the model to replicate the target transformation. Our on-the-fly supervision ensures the model to learn temporally consistent and dynamic transformations. Extensive experiments demonstrate that our PropFly significantly outperforms the state-of-the-art methods on various video editing tasks, producing high-quality editing results.

Jihyong Oh, Munchurl Kim

In this paper, we propose a novel joint deblurring and multi-frame interpolation (DeMFI) framework, called DeMFI-Net, which accurately converts blurry videos of lower-frame-rate to sharp videos at higher-frame-rate based on flow-guided attentive-correlation-based feature bolstering (FAC-FB) module and recursive boosting (RB), in terms of multi-frame interpolation (MFI). The DeMFI-Net jointly performs deblurring and MFI where its baseline version performs feature-flow-based warping with FAC-FB module to obtain a sharp-interpolated frame as well to deblur two center-input frames. Moreover, its extended version further improves the joint task performance based on pixel-flow-based warping with GRU-based RB. Our FAC-FB module effectively gathers the distributed blurry pixel information over blurry input frames in feature-domain to improve the overall joint performances, which is computationally efficient since its attentive correlation is only focused pointwise. As a result, our DeMFI-Net achieves state-of-the-art (SOTA) performances for diverse datasets with significant margins compared to the recent SOTA methods, for both deblurring and MFI. All source codes including pretrained DeMFI-Net are publicly available at https://github.com/JihyongOh/DeMFI.

Hyeonjun Sim, Jihyong Oh, Munchurl Kim

In this paper, we firstly present a dataset (X4K1000FPS) of 4K videos of 1000 fps with the extreme motion to the research community for video frame interpolation (VFI), and propose an extreme VFI network, called XVFI-Net, that first handles the VFI for 4K videos with large motion. The XVFI-Net is based on a recursive multi-scale shared structure that consists of two cascaded modules for bidirectional optical flow learning between two input frames (BiOF-I) and for bidirectional optical flow learning from target to input frames (BiOF-T). The optical flows are stably approximated by a complementary flow reversal (CFR) proposed in BiOF-T module. During inference, the BiOF-I module can start at any scale of input while the BiOF-T module only operates at the original input scale so that the inference can be accelerated while maintaining highly accurate VFI performance. Extensive experimental results show that our XVFI-Net can successfully capture the essential information of objects with extremely large motions and complex textures while the state-of-the-art methods exhibit poor performance. Furthermore, our XVFI-Net framework also performs comparably on the previous lower resolution benchmark dataset, which shows a robustness of our algorithm as well. All source codes, pre-trained models, and proposed X4K1000FPS datasets are publicly available at https://github.com/JihyongOh/XVFI.

Soo Ye Kim, Kfir Aberman, Nori Kanazawa et al.

Although deep learning has enabled a huge leap forward in image inpainting, current methods are often unable to synthesize realistic high-frequency details. In this paper, we propose applying super-resolution to coarsely reconstructed outputs, refining them at high resolution, and then downscaling the output to the original resolution. By introducing high-resolution images to the refinement network, our framework is able to reconstruct finer details that are usually smoothed out due to spectral bias - the tendency of neural networks to reconstruct low frequencies better than high frequencies. To assist training the refinement network on large upscaled holes, we propose a progressive learning technique in which the size of the missing regions increases as training progresses. Our zoom-in, refine and zoom-out strategy, combined with high-resolution supervision and progressive learning, constitutes a framework-agnostic approach for enhancing high-frequency details that can be applied to any CNN-based inpainting method. We provide qualitative and quantitative evaluations along with an ablation analysis to show the effectiveness of our approach. This seemingly simple, yet powerful approach, outperforms state-of-the-art inpainting methods. Our code is available in https://github.com/google/zoom-to-inpaint

Geunhyuk Youk, Jihyong Oh, Munchurl Kim

We present a joint learning scheme of video super-resolution and deblurring, called VSRDB, to restore clean high-resolution (HR) videos from blurry low-resolution (LR) ones. This joint restoration problem has drawn much less attention compared to single restoration problems. In this paper, we propose a novel flow-guided dynamic filtering (FGDF) and iterative feature refinement with multi-attention (FRMA), which constitutes our VSRDB framework, denoted as FMA-Net. Specifically, our proposed FGDF enables precise estimation of both spatio-temporally-variant degradation and restoration kernels that are aware of motion trajectories through sophisticated motion representation learning. Compared to conventional dynamic filtering, the FGDF enables the FMA-Net to effectively handle large motions into the VSRDB. Additionally, the stacked FRMA blocks trained with our novel temporal anchor (TA) loss, which temporally anchors and sharpens features, refine features in a course-to-fine manner through iterative updates. Extensive experiments demonstrate the superiority of the proposed FMA-Net over state-of-the-art methods in terms of both quantitative and qualitative quality. Codes and pre-trained models are available at: https://kaist-viclab.github.io/fmanet-site

Jongmin Park, Minh-Quan Viet Bui, Juan Luis Gonzalez Bello et al.

Synthesizing novel views from in-the-wild monocular videos is challenging due to scene dynamics and the lack of multi-view cues. To address this, we propose SplineGS, a COLMAP-free dynamic 3D Gaussian Splatting (3DGS) framework for high-quality reconstruction and fast rendering from monocular videos. At its core is a novel Motion-Adaptive Spline (MAS) method, which represents continuous dynamic 3D Gaussian trajectories using cubic Hermite splines with a small number of control points. For MAS, we introduce a Motion-Adaptive Control points Pruning (MACP) method to model the deformation of each dynamic 3D Gaussian across varying motions, progressively pruning control points while maintaining dynamic modeling integrity. Additionally, we present a joint optimization strategy for camera parameter estimation and 3D Gaussian attributes, leveraging photometric and geometric consistency. This eliminates the need for Structure-from-Motion preprocessing and enhances SplineGS's robustness in real-world conditions. Experiments show that SplineGS significantly outperforms state-of-the-art methods in novel view synthesis quality for dynamic scenes from monocular videos, achieving thousands times faster rendering speed.

Jaeho Moon, Juan Luis Gonzalez Bello, Byeongjun Kwon et al.

Self-supervised monocular depth estimation (DE) is an approach to learning depth without costly depth ground truths. However, it often struggles with moving objects that violate the static scene assumption during training. To address this issue, we introduce a coarse-to-fine training strategy leveraging the ground contacting prior based on the observation that most moving objects in outdoor scenes contact the ground. In the coarse training stage, we exclude the objects in dynamic classes from the reprojection loss calculation to avoid inaccurate depth learning. To provide precise supervision on the depth of the objects, we present a novel Ground-contacting-prior Disparity Smoothness Loss (GDS-Loss) that encourages a DE network to align the depth of the objects with their ground-contacting points. Subsequently, in the fine training stage, we refine the DE network to learn the detailed depth of the objects from the reprojection loss, while ensuring accurate DE on the moving object regions by employing our regularization loss with a cost-volume-based weighting factor. Our overall coarse-to-fine training strategy can easily be integrated with existing DE methods without any modifications, significantly enhancing DE performance on challenging Cityscapes and KITTI datasets, especially in the moving object regions.

Sangwoon Kwak, Joonsoo Kim, Jun Young Jeong et al.

3D Gaussian Splatting (3DGS) has made significant strides in scene representation and neural rendering, with intense efforts focused on adapting it for dynamic scenes. Despite delivering remarkable rendering quality and speed, existing methods struggle with storage demands and representing complex real-world motions. To tackle these issues, we propose MoDecGS, a memory-efficient Gaussian splatting framework designed for reconstructing novel views in challenging scenarios with complex motions. We introduce GlobaltoLocal Motion Decomposition (GLMD) to effectively capture dynamic motions in a coarsetofine manner. This approach leverages Global Canonical Scaffolds (Global CS) and Local Canonical Scaffolds (Local CS), extending static Scaffold representation to dynamic video reconstruction. For Global CS, we propose Global Anchor Deformation (GAD) to efficiently represent global dynamics along complex motions, by directly deforming the implicit Scaffold attributes which are anchor position, offset, and local context features. Next, we finely adjust local motions via the Local Gaussian Deformation (LGD) of Local CS explicitly. Additionally, we introduce Temporal Interval Adjustment (TIA) to automatically control the temporal coverage of each Local CS during training, allowing MoDecGS to find optimal interval assignments based on the specified number of temporal segments. Extensive evaluations demonstrate that MoDecGS achieves an average 70% reduction in model size over stateoftheart methods for dynamic 3D Gaussians from realworld dynamic videos while maintaining or even improving rendering quality.

Jeonghyeok Do, Jaehyup Lee, Munchurl Kim

Synthetic Aperture Radar (SAR) imagery provides robust environmental and temporal coverage (e.g., during clouds, seasons, day-night cycles), yet its noise and unique structural patterns pose interpretation challenges, especially for non-experts. SAR-to-EO (Electro-Optical) image translation (SET) has emerged to make SAR images more perceptually interpretable. However, traditional approaches trained from scratch on limited SAR-EO datasets are prone to overfitting. To address these challenges, we introduce Confidence Diffusion for SAR-to-EO Translation, called C-DiffSET, a framework leveraging pretrained Latent Diffusion Model (LDM) extensively trained on natural images, thus enabling effective adaptation to the EO domain. Remarkably, we find that the pretrained VAE encoder aligns SAR and EO images in the same latent space, even with varying noise levels in SAR inputs. To further improve pixel-wise fidelity for SET, we propose a confidence-guided diffusion (C-Diff) loss that mitigates artifacts from temporal discrepancies, such as appearing or disappearing objects, thereby enhancing structural accuracy. C-DiffSET achieves state-of-the-art (SOTA) results on multiple datasets, significantly outperforming the very recent image-to-image translation methods and SET methods with large margins.

Minh-Quan Viet Bui, Jongmin Park, Jihyong Oh et al.

Neural Radiance Fields (NeRF), initially developed for static scenes, have inspired many video novel view synthesis techniques. However, the challenge for video view synthesis arises from motion blur, a consequence of object or camera movements during exposure, which hinders the precise synthesis of sharp spatio-temporal views. In response, we propose a novel motion deblurring NeRF framework for blurry monocular video, called MoBluRF, consisting of a Base Ray Initialization (BRI) stage and a Motion Decomposition-based Deblurring (MDD) stage. In the BRI stage, we coarsely reconstruct dynamic 3D scenes and jointly initialize the base rays which are further used to predict latent sharp rays, using the inaccurate camera pose information from the given blurry frames. In the MDD stage, we introduce a novel Incremental Latent Sharp-rays Prediction (ILSP) approach for the blurry monocular video frames by decomposing the latent sharp rays into global camera motion and local object motion components. We further propose two loss functions for effective geometry regularization and decomposition of static and dynamic scene components without any mask supervision. Experiments show that MoBluRF outperforms qualitatively and quantitatively the recent state-of-the-art methods with large margins.

Juan Luis Gonzalez Bello, Munchurl Kim

In this paper, we firstly consider view-dependent effects into single image-based novel view synthesis (NVS) problems. For this, we propose to exploit the camera motion priors in NVS to model view-dependent appearance or effects (VDE) as the negative disparity in the scene. By recognizing specularities "follow" the camera motion, we infuse VDEs into the input images by aggregating input pixel colors along the negative depth region of the epipolar lines. Also, we propose a `relaxed volumetric rendering' approximation that allows computing the densities in a single pass, improving efficiency for NVS from single images. Our method can learn single-image NVS from image sequences only, which is a completely self-supervised learning method, for the first time requiring neither depth nor camera pose annotations. We present extensive experiment results and show that our proposed method can learn NVS with VDEs, outperforming the SOTA single-view NVS methods on the RealEstate10k and MannequinChallenge datasets.

Minh-Quan Viet Bui, Jongmin Park, Juan Luis Gonzalez Bello et al.

We present MoBGS, a novel deblurring dynamic 3D Gaussian Splatting (3DGS) framework capable of reconstructing sharp and high-quality novel spatio-temporal views from blurry monocular videos in an end-to-end manner. Existing dynamic novel view synthesis (NVS) methods are highly sensitive to motion blur in casually captured videos, resulting in significant degradation of rendering quality. While recent approaches address motion-blurred inputs for NVS, they primarily focus on static scene reconstruction and lack dedicated motion modeling for dynamic objects. To overcome these limitations, our MoBGS introduces a novel Blur-adaptive Latent Camera Estimation (BLCE) method for effective latent camera trajectory estimation, improving global camera motion deblurring. In addition, we propose a physically-inspired Latent Camera-induced Exposure Estimation (LCEE) method to ensure consistent deblurring of both global camera and local object motion. Our MoBGS framework ensures the temporal consistency of unseen latent timestamps and robust motion decomposition of static and dynamic regions. Extensive experiments on the Stereo Blur dataset and real-world blurry videos show that our MoBGS significantly outperforms the very recent advanced methods (DyBluRF and Deblur4DGS), achieving state-of-the-art performance for dynamic NVS under motion blur.

Sungpyo Kim, Jeonghyeok Do, Jaehyup Lee et al.

Conventional methods for PAN-sharpening often struggle to restore fine details due to limitations in leveraging high-frequency information. Moreover, diffusion-based approaches lack sufficient conditioning to fully utilize Panchromatic (PAN) images and low-resolution multispectral (LRMS) inputs effectively. To address these challenges, we propose an uncertainty-aware knowledge distillation diffusion framework with details enhancement for PAN-sharpening, called U-Know-DiffPAN. The U-Know-DiffPAN incorporates uncertainty-aware knowledge distillation for effective transfer of feature details from our teacher model to a student one. The teacher model in our U-Know-DiffPAN captures frequency details through freqeuncy selective attention, facilitating accurate reverse process learning. By conditioning the encoder on compact vector representations of PAN and LRMS and the decoder on Wavelet transforms, we enable rich frequency utilization. So, the high-capacity teacher model distills frequency-rich features into a lightweight student model aided by an uncertainty map. From this, the teacher model can guide the student model to focus on difficult image regions for PAN-sharpening via the usage of the uncertainty map. Extensive experiments on diverse datasets demonstrate the robustness and superior performance of our U-Know-DiffPAN over very recent state-of-the-art PAN-sharpening methods.

Wonyong Seo, Jihyong Oh, Munchurl Kim

Existing Video Frame interpolation (VFI) models tend to suffer from time-to-location ambiguity when trained with video of non-uniform motions, such as accelerating, decelerating, and changing directions, which often yield blurred interpolated frames. In this paper, we propose (i) a novel motion description map, Bidirectional Motion field (BiM), to effectively describe non-uniform motions; (ii) a BiM-guided Flow Net (BiMFN) with Content-Aware Upsampling Network (CAUN) for precise optical flow estimation; and (iii) Knowledge Distillation for VFI-centric Flow supervision (KDVCF) to supervise the motion estimation of VFI model with VFI-centric teacher flows. The proposed VFI is called a Bidirectional Motion field-guided VFI (BiM-VFI) model. Extensive experiments show that our BiM-VFI model significantly surpasses the recent state-of-the-art VFI methods by 26% and 45% improvements in LPIPS and STLPIPS respectively, yielding interpolated frames with much fewer blurs at arbitrary time instances.

Jeonghyeok Do, Munchurl Kim

In zero-shot skeleton-based action recognition (ZSAR), aligning skeleton features with the text features of action labels is essential for accurately predicting unseen actions. ZSAR faces a fundamental challenge in bridging the modality gap between the two-kind features, which severely limits generalization to unseen actions. Previous methods focus on direct alignment between skeleton and text latent spaces, but the modality gaps between these spaces hinder robust generalization learning. Motivated by the success of diffusion models in multi-modal alignment (e.g., text-to-image, text-to-video), we firstly present a diffusion-based skeleton-text alignment framework for ZSAR. Our approach, Triplet Diffusion for Skeleton-Text Matching (TDSM), focuses on cross-alignment power of diffusion models rather than their generative capability. Specifically, TDSM aligns skeleton features with text prompts by incorporating text features into the reverse diffusion process, where skeleton features are denoised under text guidance, forming a unified skeleton-text latent space for robust matching. To enhance discriminative power, we introduce a triplet diffusion (TD) loss that encourages our TDSM to correct skeleton-text matches while pushing them apart for different action classes. Our TDSM significantly outperforms very recent state-of-the-art methods with significantly large margins of 2.36%-point to 13.05%-point, demonstrating superior accuracy and scalability in zero-shot settings through effective skeleton-text matching.

Samuel Teodoro, Yun Chen, Agus Gunawan et al.

Motion transfer enables controllable video generation by transferring temporal dynamics from a reference video to synthesize a new video conditioned on a target caption. However, existing Diffusion Transformer (DiT)-based methods are limited to single-object videos, restricting fine-grained control in real-world scenes with multiple objects. In this work, we introduce MotionGrounder, a DiT-based framework that firstly handles motion transfer with multi-object controllability. Our Flow-based Motion Signal (FMS) in MotionGrounder provides a stable motion prior for target video generation, while our Object-Caption Alignment Loss (OCAL) grounds object captions to their corresponding spatial regions. We further propose a new Object Grounding Score (OGS), which jointly evaluates (i) spatial alignment between source video objects and their generated counterparts and (ii) semantic consistency between each generated object and its target caption. Our experiments show that MotionGrounder consistently outperforms recent baselines across quantitative, qualitative, and human evaluations.

Agus Gunawan, Samuel Teodoro, Yun Chen et al.

Recent advancements in diffusion-based text synthesis have demonstrated significant performance in inserting and editing text within images via inpainting. However, despite the potential of text inpainting methods, three key limitations hinder their applicability to broader Text Image Manipulation (TIM) tasks: (i) the inability to remove text, (ii) the lack of control over the style of rendered text, and (iii) a tendency to generate duplicated letters. To address these challenges, we propose OmniText, a training-free generalist capable of performing a wide range of TIM tasks. Specifically, we investigate two key properties of cross- and self-attention mechanisms to enable text removal and to provide control over both text styles and content. Our findings reveal that text removal can be achieved by applying self-attention inversion, which mitigates the model's tendency to focus on surrounding text, thus reducing text hallucinations. Additionally, we redistribute cross-attention, as increasing the probability of certain text tokens reduces text hallucination. For controllable inpainting, we introduce novel loss functions in a latent optimization framework: a cross-attention content loss to improve text rendering accuracy and a self-attention style loss to facilitate style customization. Furthermore, we present OmniText-Bench, a benchmark dataset for evaluating diverse TIM tasks. It includes input images, target text with masks, and style references, covering diverse applications such as text removal, rescaling, repositioning, and insertion and editing with various styles. Our OmniText framework is the first generalist method capable of performing diverse TIM tasks. It achieves state-of-the-art performance across multiple tasks and metrics compared to other text inpainting methods and is comparable with specialist methods.

Jeonghyeok Do, Sungpyo Kim, Geunhyuk Youk et al.

PAN-sharpening aims to fuse high-resolution panchromatic (PAN) images with low-resolution multi-spectral (MS) images to generate high-resolution multi-spectral (HRMS) outputs. However, cross-modality misalignment -- caused by sensor placement, acquisition timing, and resolution disparity -- induces a fundamental challenge. Conventional deep learning methods assume perfect pixel-wise alignment and rely on per-pixel reconstruction losses, leading to spectral distortion, double edges, and blurring when misalignment is present. To address this, we propose PAN-Crafter, a modality-consistent alignment framework that explicitly mitigates the misalignment gap between PAN and MS modalities. At its core, Modality-Adaptive Reconstruction (MARs) enables a single network to jointly reconstruct HRMS and PAN images, leveraging PAN's high-frequency details as auxiliary self-supervision. Additionally, we introduce Cross-Modality Alignment-Aware Attention (CM3A), a novel mechanism that bidirectionally aligns MS texture to PAN structure and vice versa, enabling adaptive feature refinement across modalities. Extensive experiments on multiple benchmark datasets demonstrate that our PAN-Crafter outperforms the most recent state-of-the-art method in all metrics, even with 50.11$\times$ faster inference time and 0.63$\times$ the memory size. Furthermore, it demonstrates strong generalization performance on unseen satellite datasets, showing its robustness across different conditions.

Byeongjun Kwon, Munchurl Kim

Zero-shot depth estimation (DE) models exhibit strong generalization performance as they are trained on large-scale datasets. However, existing models struggle with high-resolution images due to the discrepancy in image resolutions of training (with smaller resolutions) and inference (for high resolutions). Processing them at full resolution leads to decreased estimation accuracy on depth with tremendous memory consumption, while downsampling to the training resolution results in blurred edges in the estimated depth images. Prevailing high-resolution depth estimation methods adopt a patch-based approach, which introduces depth discontinuity issues when reassembling the estimated depth patches, resulting in test-time inefficiency. Additionally, to obtain fine-grained depth details, these methods rely on synthetic datasets due to the real-world sparse ground truth depth, leading to poor generalizability. To tackle these limitations, we propose Patch Refine Once (PRO), an efficient and generalizable tile-based framework. Our PRO consists of two key components: (i) Grouped Patch Consistency Training that enhances test-time efficiency while mitigating the depth discontinuity problem by jointly processing four overlapping patches and enforcing a consistency loss on their overlapping regions within a single backpropagation step, and (ii) Bias Free Masking that prevents the DE models from overfitting to dataset-specific biases, enabling better generalization to real-world datasets even after training on synthetic data. Zero-shot evaluations on Booster, ETH3D, Middlebury 2014, and NuScenes demonstrate that our PRO can be seamlessly integrated into existing depth estimation models.

Samuel Teodoro, Agus Gunawan, Soo Ye Kim et al.

Recent AI-based video editing has enabled users to edit videos through simple text prompts, significantly simplifying the editing process. However, recent zero-shot video editing techniques primarily focus on global or single-object edits, which can lead to unintended changes in other parts of the video. When multiple objects require localized edits, existing methods face challenges, such as unfaithful editing, editing leakage, and lack of suitable evaluation datasets and metrics. To overcome these limitations, we propose $\textbf{P}$robability $\textbf{R}$edistribution for $\textbf{I}$nstance-aware $\textbf{M}$ulti-object Video $\textbf{Edit}$ing ($\textbf{PRIMEdit}$). PRIMEdit is a zero-shot framework that introduces two key modules: (i) Instance-centric Probability Redistribution (IPR) to ensure precise localization and faithful editing and (ii) Disentangled Multi-instance Sampling (DMS) to prevent editing leakage. Additionally, we present our new MIVE Dataset for video editing featuring diverse video scenarios, and introduce the Cross-Instance Accuracy (CIA) Score to evaluate editing leakage in multi-instance video editing tasks. Our extensive qualitative, quantitative, and user study evaluations demonstrate that PRIMEdit significantly outperforms recent state-of-the-art methods in terms of editing faithfulness, accuracy, and leakage prevention, setting a new benchmark for multi-instance video editing.

Jaeho Moon, Jeonghwan Yun, Jaehyun Kim et al.

Oil spills pose severe environmental risks, making early detection crucial for effective response and mitigation. As Synthetic Aperture Radar (SAR) images operate under all-weather conditions, SAR-based oil spill segmentation enables fast and robust monitoring. However, when using deep learning models, SAR oil spill segmentation often struggles in training due to the scarcity of labeled data. To address this limitation, we propose a diffusion-based data augmentation with knowledge transfer (DAKTer) strategy. Our DAKTer strategy enables a diffusion model to generate SAR oil spill images along with soft label pairs, which offer richer class probability distributions than segmentation masks (i.e. hard labels). Also, for reliable joint generation of high-quality SAR images and well-aligned soft labels, we introduce an SNR-based balancing factor aligning the noise corruption process of both modalilties in diffusion models. By leveraging the generated SAR images and soft labels, a student segmentation model can learn robust feature representations without teacher models trained for the same task, improving its ability to segment oil spill regions. Extensive experiments demonstrate that our DAKTer strategy effectively transfers the knowledge of per-pixel class probabilities to the student segmentation model to distinguish the oil spill regions from other look-alike regions in the SAR images. Our DAKTer strategy boosts various segmentation models to achieve superior performance with large margins compared to other generative data augmentation methods.

Jeonghyeok Do, Munchurl Kim

Skeleton-based action recognition, which classifies human actions based on the coordinates of joints and their connectivity within skeleton data, is widely utilized in various scenarios. While Graph Convolutional Networks (GCNs) have been proposed for skeleton data represented as graphs, they suffer from limited receptive fields constrained by joint connectivity. To address this limitation, recent advancements have introduced transformer-based methods. However, capturing correlations between all joints in all frames requires substantial memory resources. To alleviate this, we propose a novel approach called Skeletal-Temporal Transformer (SkateFormer) that partitions joints and frames based on different types of skeletal-temporal relation (Skate-Type) and performs skeletal-temporal self-attention (Skate-MSA) within each partition. We categorize the key skeletal-temporal relations for action recognition into a total of four distinct types. These types combine (i) two skeletal relation types based on physically neighboring and distant joints, and (ii) two temporal relation types based on neighboring and distant frames. Through this partition-specific attention strategy, our SkateFormer can selectively focus on key joints and frames crucial for action recognition in an action-adaptive manner with efficient computation. Extensive experiments on various benchmark datasets validate that our SkateFormer outperforms recent state-of-the-art methods.

Juan Luis Gonzalez Bello, Minh-Quan Viet Bui, Munchurl Kim

Recent advances in neural rendering have shown that, albeit slow, implicit compact models can learn a scene's geometries and view-dependent appearances from multiple views. To maintain such a small memory footprint but achieve faster inference times, recent works have adopted `sampler' networks that adaptively sample a small subset of points along each ray in the implicit neural radiance fields. Although these methods achieve up to a 10$\times$ reduction in rendering time, they still suffer from considerable quality degradation compared to the vanilla NeRF. In contrast, we propose ProNeRF, which provides an optimal trade-off between memory footprint (similar to NeRF), speed (faster than HyperReel), and quality (better than K-Planes). ProNeRF is equipped with a novel projection-aware sampling (PAS) network together with a new training strategy for ray exploration and exploitation, allowing for efficient fine-grained particle sampling. Our ProNeRF yields state-of-the-art metrics, being 15-23x faster with 0.65dB higher PSNR than NeRF and yielding 0.95dB higher PSNR than the best published sampler-based method, HyperReel. Our exploration and exploitation training strategy allows ProNeRF to learn the full scenes' color and density distributions while also learning efficient ray sampling focused on the highest-density regions. We provide extensive experimental results that support the effectiveness of our method on the widely adopted forward-facing and 360 datasets, LLFF and Blender, respectively.

Jaehyup Lee, Soomin Seo, Munchurl Kim

Pan-sharpening is a process of merging a high-resolution (HR) panchromatic (PAN) image and its corresponding low-resolution (LR) multi-spectral (MS) image to create an HR-MS and pan-sharpened image. However, due to the different sensors' locations, characteristics and acquisition time, PAN and MS image pairs often tend to have various amounts of misalignment. Conventional deep-learning-based methods that were trained with such misaligned PAN-MS image pairs suffer from diverse artifacts such as double-edge and blur artifacts in the resultant PAN-sharpened images. In this paper, we propose a novel framework called shift-invariant pan-sharpening with moving object alignment (SIPSA-Net) which is the first method to take into account such large misalignment of moving object regions for PAN sharpening. The SISPA-Net has a feature alignment module (FAM) that can adjust one feature to be aligned to another feature, even between the two different PAN and MS domains. For better alignment in pan-sharpened images, a shift-invariant spectral loss is newly designed, which ignores the inherent misalignment in the original MS input, thereby having the same effect as optimizing the spectral loss with a well-aligned MS image. Extensive experimental results show that our SIPSA-Net can generate pan-sharpened images with remarkable improvements in terms of visual quality and alignment, compared to the state-of-the-art methods.

Dac Tung Vu, Juan Luis Gonzalez, Munchurl Kim

Heavy rain removal from a single image is the task of simultaneously eliminating rain streaks and fog, which can dramatically degrade the quality of captured images. Most existing rain removal methods do not generalize well for the heavy rain case. In this work, we propose a novel network architecture consisting of three sub-networks to remove heavy rain from a single image without estimating rain streaks and fog separately. The first sub-net, a U-net-based architecture that incorporates our Spatial Channel Attention (SCA) blocks, extracts global features that provide sufficient contextual information needed to remove atmospheric distortions caused by rain and fog. The second sub-net learns the additive residues information, which is useful in removing rain streak artifacts via our proposed Residual Inception Modules (RIM). The third sub-net, the multiplicative sub-net, adopts our Channel-attentive Inception Modules (CIM) and learns the essential brighter local features which are not effectively extracted in the SCA and additive sub-nets by modulating the local pixel intensities in the derained images. Our three clean image results are then combined via an attentive blending block to generate the final clean image. Our method with SCA, RIM, and CIM significantly outperforms the previous state-of-the-art single-image deraining methods on the synthetic datasets, shows considerably cleaner and sharper derained estimates on the real image datasets. We present extensive experiments and ablation studies supporting each of our method's contributions on both synthetic and real image datasets.

Jihyong Oh, Munchurl Kim

Although Generative Adversarial Networks (GANs) are successfully applied to diverse fields, training GANs on synthetic aperture radar (SAR) data is a challenging task mostly due to speckle noise. On the one hands, in a learning perspective of human's perception, it is natural to learn a task by using various information from multiple sources. However, in the previous GAN works on SAR target image generation, the information on target classes has only been used. Due to the backscattering characteristics of SAR image signals, the shapes and structures of SAR target images are strongly dependent on their pose angles. Nevertheless, the pose angle information has not been incorporated into such generative models for SAR target images. In this paper, we firstly propose a novel GAN-based multi-task learning (MTL) method for SAR target image generation, called PeaceGAN that uses both pose angle and target class information, which makes it possible to produce SAR target images of desired target classes at intended pose angles. For this, the PeaceGAN has two additional structures, a pose estimator and an auxiliary classifier, at the side of its discriminator to combine the pose and class information more efficiently. In addition, the PeaceGAN is jointly learned in an end-to-end manner as MTL with both pose angle and target class information, thus enhancing the diversity and quality of generated SAR target images The extensive experiments show that taking an advantage of both pose angle and target class learning by the proposed pose estimator and auxiliary classifier can help the PeaceGAN's generator effectively learn the distributions of SAR target images in the MTL framework, so that it can better generate the SAR target images more flexibly and faithfully at intended pose angles for desired target classes compared to the recent state-of-the-art methods.

Juan Luis Gonzalez Bello, Munchurl Kim

In this paper, we propose a self-supervised single-view pixel-level accurate depth estimation network, called PLADE-Net. The PLADE-Net is the first work that shows unprecedented accuracy levels, exceeding 95\% in terms of the $δ^1$ metric on the challenging KITTI dataset. Our PLADE-Net is based on a new network architecture with neural positional encoding and a novel loss function that borrows from the closed-form solution of the matting Laplacian to learn pixel-level accurate depth estimation from stereo images. Neural positional encoding allows our PLADE-Net to obtain more consistent depth estimates by letting the network reason about location-specific image properties such as lens and projection distortions. Our novel distilled matting Laplacian loss allows our network to predict sharp depths at object boundaries and more consistent depths in highly homogeneous regions. Our proposed method outperforms all previous self-supervised single-view depth estimation methods by a large margin on the challenging KITTI dataset, with unprecedented levels of accuracy. Furthermore, our PLADE-Net, naively extended for stereo inputs, outperforms the most recent self-supervised stereo methods, even without any advanced blocks like 1D correlations, 3D convolutions, or spatial pyramid pooling. We present extensive ablation studies and experiments that support our method's effectiveness on the KITTI, CityScapes, and Make3D datasets.

Soo Ye Kim, Hyeonjun Sim, Munchurl Kim

Blind super-resolution (SR) methods aim to generate a high quality high resolution image from a low resolution image containing unknown degradations. However, natural images contain various types and amounts of blur: some may be due to the inherent degradation characteristics of the camera, but some may even be intentional, for aesthetic purposes (e.g. Bokeh effect). In the case of the latter, it becomes highly difficult for SR methods to disentangle the blur to remove, and that to leave as is. In this paper, we propose a novel blind SR framework based on kernel-oriented adaptive local adjustment (KOALA) of SR features, called KOALAnet, which jointly learns spatially-variant degradation and restoration kernels in order to adapt to the spatially-variant blur characteristics in real images. Our KOALAnet outperforms recent blind SR methods for synthesized LR images obtained with randomized degradations, and we further show that the proposed KOALAnet produces the most natural results for artistic photographs with intentional blur, which are not over-sharpened, by effectively handling images mixed with in-focus and out-of-focus areas.

Juan Luis Gonzalez, Munchurl Kim

Self-supervised depth estimators have recently shown results comparable to the supervised methods on the challenging single image depth estimation (SIDE) task, by exploiting the geometrical relations between target and reference views in the training data. However, previous methods usually learn forward or backward image synthesis, but not depth estimation, as they cannot effectively neglect occlusions between the target and the reference images. Previous works rely on rigid photometric assumptions or the SIDE network to infer depth and occlusions, resulting in limited performance. On the other hand, we propose a method to "Forget About the LiDAR" (FAL), for the training of depth estimators, with Mirrored Exponential Disparity (MED) probability volumes, from which we obtain geometrically inspired occlusion maps with our novel Mirrored Occlusion Module (MOM). Our MOM does not impose a burden on our FAL-net. Contrary to the previous methods that learn SIDE from stereo pairs by regressing disparity in the linear space, our FAL-net regresses disparity by binning it into the exponential space, which allows for better detection of distant and nearby objects. We define a two-step training strategy for our FAL-net: It is first trained for view synthesis and then fine-tuned for depth estimation with our MOM. Our FAL-net is remarkably light-weight and outperforms the previous state-of-the-art methods with 8x fewer parameters and 3x faster inference speeds on the challenging KITTI dataset. We present extensive experimental results on the KITTI, CityScapes, and Make3D datasets to verify our method's effectiveness. To the authors' best knowledge, the presented method performs the best among all the previous self-supervised methods until now.

Codruta O. Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu et al.

This paper reviews the NTIRE 2020 Challenge on NonHomogeneous Dehazing of images (restoration of rich details in hazy image). We focus on the proposed solutions and their results evaluated on NH-Haze, a novel dataset consisting of 55 pairs of real haze free and nonhomogeneous hazy images recorded outdoor. NH-Haze is the first realistic nonhomogeneous haze dataset that provides ground truth images. The nonhomogeneous haze has been produced using a professional haze generator that imitates the real conditions of haze scenes. 168 participants registered in the challenge and 27 teams competed in the final testing phase. The proposed solutions gauge the state-of-the-art in image dehazing.

Jooyoung Lee, Seunghyun Cho, Munchurl Kim

Recently, learned image compression methods have been actively studied. Among them, entropy-minimization based approaches have achieved superior results compared to conventional image codecs such as BPG and JPEG2000. However, the quality enhancement and rate-minimization are conflictively coupled in the process of image compression. That is, maintaining high image quality entails less compression and vice versa. However, by jointly training separate quality enhancement in conjunction with image compression, the coding efficiency can be improved. In this paper, we propose a novel joint learning scheme of image compression and quality enhancement, called JointIQ-Net, as well as entropy model improvement, thus achieving significantly improved coding efficiency against the previous methods. Our proposed JointIQ-Net combines an image compression sub-network and a quality enhancement sub-network in a cascade, both of which are end-to-end trained in a combined manner within the JointIQ-Net. Also the JointIQ-Net benefits from improved entropy-minimization that newly adopts a Gussian Mixture Model (GMM) and further exploits global context to estimate the probabilities of latent representations. In order to show the effectiveness of our proposed JointIQ-Net, extensive experiments have been performed, and showed that the JointIQ-Net achieves a remarkable performance improvement in coding efficiency in terms of both PSNR and MS-SSIM, compared to the previous learned image compression methods and the conventional codecs such as VVC Intra (VTM 7.1), BPG, and JPEG2000. To the best of our knowledge, this is the first end-to-end optimized image compression method that outperforms VTM 7.1 (Intra), the latest reference software of the VVC standard, in terms of the PSNR and MS-SSIM.

Soo Ye Kim, Jihyong Oh, Munchurl Kim

Super-resolution (SR) has been widely used to convert low-resolution legacy videos to high-resolution (HR) ones, to suit the increasing resolution of displays (e.g. UHD TVs). However, it becomes easier for humans to notice motion artifacts (e.g. motion judder) in HR videos being rendered on larger-sized display devices. Thus, broadcasting standards support higher frame rates for UHD (Ultra High Definition) videos (4K@60 fps, 8K@120 fps), meaning that applying SR only is insufficient to produce genuine high quality videos. Hence, to up-convert legacy videos for realistic applications, not only SR but also video frame interpolation (VFI) is necessitated. In this paper, we first propose a joint VFI-SR framework for up-scaling the spatio-temporal resolution of videos from 2K 30 fps to 4K 60 fps. For this, we propose a novel training scheme with a multi-scale temporal loss that imposes temporal regularization on the input video sequence, which can be applied to any general video-related task. The proposed structure is analyzed in depth with extensive experiments.

Juan Luis Gonzalez Bello, Munchurl Kim

Recent advances in deep learning have shown promising results in many low-level vision tasks. However, solving the single-image-based view synthesis is still an open problem. In particular, the generation of new images at parallel camera views given a single input image is of great interest, as it enables 3D visualization of the 2D input scenery. We propose a novel network architecture to perform stereoscopic view synthesis at arbitrary camera positions along the X-axis, or Deep 3D Pan, with "t-shaped" adaptive kernels equipped with globally and locally adaptive dilations. Our proposed network architecture, the monster-net, is devised with a novel "t-shaped" adaptive kernel with globally and locally adaptive dilation, which can efficiently incorporate global camera shift into and handle local 3D geometries of the target image's pixels for the synthesis of naturally looking 3D panned views when a 2-D input image is given. Extensive experiments were performed on the KITTI, CityScapes and our VICLAB_STEREO indoors dataset to prove the efficacy of our method. Our monster-net significantly outperforms the state-of-the-art method, SOTA, by a large margin in all metrics of RMSE, PSNR, and SSIM. Our proposed monster-net is capable of reconstructing more reliable image structures in synthesized images with coherent geometry. Moreover, the disparity information that can be extracted from the "t-shaped" kernel is much more reliable than that of the SOTA for the unsupervised monocular depth estimation task, confirming the effectiveness of our method.

Juan Luis Gonzalez Bello, Munchurl Kim

The 3D-zoom operation is the positive translation of the camera in the Z-axis, perpendicular to the image plane. In contrast, the optical zoom changes the focal length and the digital zoom is used to enlarge a certain region of an image to the original image size. In this paper, we are the first to formulate an unsupervised 3D-zoom learning problem where images with an arbitrary zoom factor can be generated from a given single image. An unsupervised framework is convenient, as it is a challenging task to obtain a 3D-zoom dataset of natural scenes due to the need for special equipment to ensure camera movement is restricted to the Z-axis. In addition, the objects in the scenes should not move when being captured, which hinders the construction of a large dataset of outdoor scenes. We present a novel unsupervised framework to learn how to generate arbitrarily 3D-zoomed versions of a single image, not requiring a 3D-zoom ground truth, called the Deep 3D-Zoom Net. The Deep 3D-Zoom Net incorporates the following features: (i) transfer learning from a pre-trained disparity estimation network via a back re-projection reconstruction loss; (ii) a fully convolutional network architecture that models depth-image-based rendering (DIBR), taking into account high-frequency details without the need for estimating the intermediate disparity; and (iii) incorporating a discriminator network that acts as a no-reference penalty for unnaturally rendered areas. Even though there is no baseline to fairly compare our results, our method outperforms previous novel view synthesis research in terms of realistic appearance on large camera baselines. We performed extensive experiments to verify the effectiveness of our method on the KITTI and Cityscapes datasets.

Soo Ye Kim, Jihyong Oh, Munchurl Kim

Joint learning of super-resolution (SR) and inverse tone-mapping (ITM) has been explored recently, to convert legacy low resolution (LR) standard dynamic range (SDR) videos to high resolution (HR) high dynamic range (HDR) videos for the growing need of UHD HDR TV/broadcasting applications. However, previous CNN-based methods directly reconstruct the HR HDR frames from LR SDR frames, and are only trained with a simple L2 loss. In this paper, we take a divide-and-conquer approach in designing a novel GAN-based joint SR-ITM network, called JSI-GAN, which is composed of three task-specific subnets: an image reconstruction subnet, a detail restoration (DR) subnet and a local contrast enhancement (LCE) subnet. We delicately design these subnets so that they are appropriately trained for the intended purpose, learning a pair of pixel-wise 1D separable filters via the DR subnet for detail restoration and a pixel-wise 2D local filter by the LCE subnet for contrast enhancement. Moreover, to train the JSI-GAN effectively, we propose a novel detail GAN loss alongside the conventional GAN loss, which helps enhancing both local details and contrasts to reconstruct high quality HR HDR results. When all subnets are jointly trained well, the predicted HR HDR results of higher quality are obtained with at least 0.41 dB gain in PSNR over those generated by the previous methods.

Jae-Seok Choi, Yongwoo Kim, Munchurl Kim

Recently, many deep-learning-based pan-sharpening methods have been proposed for generating high-quality pan-sharpened (PS) satellite images. These methods focused on various types of convolutional neural network (CNN) structures, which were trained by simply minimizing a spectral loss between network outputs and the corresponding high-resolution multi-spectral (MS) target images. However, due to different sensor characteristics and acquisition times, high-resolution panchromatic (PAN) and low-resolution MS image pairs tend to have large pixel misalignments, especially for moving objects in the images. Conventional CNNs trained with only the spectral loss with these satellite image datasets often produce PS images of low visual quality including double-edge artifacts along strong edges and ghosting artifacts on moving objects. In this letter, we propose a novel loss function, called a spectral-spatial structure (S3) loss, based on the correlation maps between MS targets and PAN inputs. Our proposed S3 loss can be very effectively utilized for pan-sharpening with various types of CNN structures, resulting in significant visual improvements on PS images with suppressed artifacts.