Je-Won Kang

Multispectral (MS) imaging extends beyond conventional RGB imaging by capturing more spectral bands, thereby improving illuminant spectrum estimation (ISE). However, existing methods often fail to fully exploit spectral information, resulting in suboptimal performance under diverse lighting conditions and across different sensor domains. Hence, we propose a deep learning framework with a spatio-spectral feature extraction block, which incorporates spectral attention mechanisms to enhance spectral correlation and preserve illuminant-relevant spatial features. Through the inclusion of an illuminant prior (IP), our approach prioritizes specific channels that provide more meaningful information in an MS image. We also propose a spectral-domain transform across different MS sensor spaces. The results demonstrate that illuminant spectra learned in high-dimensional sensor spaces can be effectively transformed to various lower-dimensional camera sensor spaces without any additional training. To facilitate evaluation, we introduce a real-world MS dataset containing high-dimensional ground-truth illumination spectra captured under diverse lighting conditions. Through extensive experiments, we demonstrate that our method achieves superior accuracy compared to existing models, thus providing a practical solution for real-world ISE. The code and dataset are available at https://github.com/hyejin5/Spectrum-Aware-Illumination-Estimation-Using-Multispectral-Image.

Green learning (GL) promotes sustainability in deep learning by emphasizing energy-efficient solutions and lightweight models. Implicit neural representations (INRs) for videos provide a compact and efficient approach to video representation within this paradigm. This study introduces SNeRV+, a spatiotemporal, spectra-preserving neural representation for a video that employs neural tangent kernel (NTK) analysis to enhance learning. To mitigate spectral bias in both spatial and temporal domains, SNeRV+ employs a two-level processing approach, where separate encoder branches handle low-frequency (LF) and high-frequency (HF) components. A three-dimensional discrete wavelet transform decomposes each frame into its temporal variations, encoding LF and HF components into frame-wise embeddings. LF components, which capture static scenes and steady motion, are decoded with fixed parameters across frames, reducing temporal discrepancies and mitigating spectral bias. HF components, which encode time-varying details, are dynamically reconstructed using temporally adaptive weights that leverage LF-related parameters as prior information. This design enables a more efficient and compact representation of temporal variations. Experimental results demonstrate that SNeRV+ outperforms state-of-the-art INR-based methods in video regression, interpolation, extrapolation, and compression, achieving superior performance across both quantitative and qualitative evaluation metrics.

Conventional unsupervised domain adaptation (UDA) requires access to source data and/or source model parameters, prohibiting its practical application in terms of privacy, security, and intellectual property. Recent black-box UDA (BDA) reduces such constraints by defining a pseudo label from a single encapsulated source application programming interface (API) prediction, which allows for self-training of the target model. Nonetheless, existing methods have limited consideration for multi-source settings, in which multiple source domain APIs are available to generate pseudo labels. In this work, we introduce a novel training framework for multi-source BDA (MSBDA), dubbed Label Space-Induced Pseudo Label Refinement (LPR). Specifically, LPR incorporates a Pseudo label Refinery Network (PRN) that learns the relationship among source domains conditioned by the target domain only utilizing source API’s prediction. The target model is adapted by our dual phases PRN. First, a warm-up phase targets to avoid failure due to noisy samples and provide an initial pseudo-label, which is followed by a label refinement phase with domain relationship exploration. We provide theoretical support for the mechanism of the LPR. Experimental results on four benchmark datasets demonstrate that MSBDA using LPR achieves competitive performance compared to state-of-the-art approaches with different DA settings.

This paper proposes a novel multi-view (MV) video coding technique that leverages a four-dimensional (4D) voxel-grid representation to enhance coding efficiency, particularly in novel view synthesis. Although the voxel grid approximation provides a continuous representation for dynamic scenes, its volumetric nature requires substantial storage. The compression of MV videos can be interpreted as the compression of dense features. However, the substantial size of these features poses a significant problem relative to the generation of dynamic scenes at arbitrary viewpoints. To address this challenge, this study introduces a hierarchical coded representation of dynamic volumes based on low-rank tensor decomposition of volumetric features and develops effective coding techniques based on this representation. The proposed method employs a two-level coding strategy to capture the temporal characteristics of the decomposed features. At a higher level, spatial features are encoded, representing 3D structural information, with time-invariant components over short intervals of an MV video sequence. At a lower level, temporal features are encoded to capture the dynamics of current scenes. The spatial features are shared in a group, and temporal features are encoded at each time step. The experimental results demonstrate that the proposed technique outperforms existing MV video coding standards and current state-of-the-art methods, providing superior rate-distortion performance in the novel view synthesis of MV video compression.

Neural representation for video (NeRV), which employs a neural network to parameterize video signals, introduces a novel methodology in video representations. However, existing NeRV-based methods have difficulty in capturing fine spatial details and motion patterns due to spectral bias, in which a neural network learns high-frequency (HF) components at a slower rate than low-frequency (LF) components. In this paper, we propose spectra-preserving NeRV (SNeRV) as a novel approach to enhance implicit video representations by efficiently handling various frequency components. SNeRV uses 2D discrete wavelet transform (DWT) to decompose video into LF and HF features, preserving spatial structures and directly addressing the spectral bias issue. To balance the compactness, we encode only the LF components, while HF components that include fine textures are generated by a decoder. Specialized modules, including a multi-resolution fusion unit (MFU) and a high-frequency restorer (HFR), are integrated into a backbone to facilitate the representation. Furthermore, we extend SNeRV to effectively capture temporal correlations between adjacent video frames, by casting the extension as additional frequency decomposition to a temporal domain. This approach allows us to embed spatio-temporal LF features into the network, using temporally extended up-sampling blocks (TUBs). Experimental results demonstrate that SNeRV outperforms existing NeRV models in capturing fine details and achieves enhanced reconstruction, making it a promising approach in the field of implicit video representations. The codes are available at https://github.com/qwertja/SNeRV.

In this article, we propose an efficient reference-based deep in-loop filtering method for video coding. Existing reference-based in-loop filters often face challenges in improving coding efficiency due to the difficulty in capturing relevant textures from the reference frames. Our method accurately predicts the texture of a reference block and uses this information to restore the current block. To achieve this, we develop a reference-to-current feature estimation module that conveys high-quality information from previously coded frames in the feature domain, thereby preventing loss of detail due to inaccurate prediction. Although a neural network is trained to restore a coded video frame to be similar to the current frame, their performance can significantly degrade when operating with various quantization parameters (QPs) and managing different levels of distortion. This problem becomes further severe in the reference-to-current feature estimation, in which QP values are applied differently to video frames. We address this problem by developing a QP-aware convolution layer with a small number of learnable parameters to generate reliable features and adapt to fine-grained adaptive QPs among consecutive frames. The proposed method is implemented into the versatile video coding (VVC) reference software, VTM version 10.0. Experimental results demonstrate that the proposed method improves coding performance significantly in VVC.