arXiv subjects

Chanung Park

Publications and source records attributed to Chanung Park.

4 recordsLinked to original sources

arXiv2026

ACA-GS: Adaptive-Capacity Anchored Gaussian Splatting for Compact Dynamic Radiance Fields

Recent advances in 4D Gaussian Splatting (4DGS) enable high-fidelity, real-time spatiotemporal rendering, but expose a fundamental trade-off between motion expressiveness and storage efficiency. While anchor-based designs achieve compactness through anchor-level parameter sharing, their rigid uniform parametrization enforces fixed Neural Gaussian counts and feature budgets per anchor. Consequently, insufficient fidelity is addressed by excessive anchor density, rather than lightweight, targeted increases in Neural Gaussian count or feature capacity, resulting in memory waste. To overcome this rigidity, we introduce an adaptive-capacity anchor-based framework that dynamically allocates the representational capacity based on local spatiotemporal demands. Adaptive Anchor Cardinality varies the number of Neural Gaussians per anchor, concentrating primitives in regions of high geometric or motion complexity while suppressing redundancy. In parallel, Adaptive Anchor Feature Masking modulates anchor-level feature channels, assigning rich features to complex regions and lightweight representations to simpler ones. Experiments on MPEG, Panoptic Sports, and N3DV datasets demonstrate substantial storage reduction without degrading visual quality. Notably, on challenging MPEG sequences with complex motion, our method achieves up to 1.5x higher compression than state-of-the-art anchor-based methods while preserving comparable quality.

cs.CV↗

arXiv2025

Single-step Diffusion for Image Compression at Ultra-Low Bitrates

Although there have been significant advancements in image compression techniques, such as standard and learned codecs, these methods still suffer from severe quality degradation at extremely low bits per pixel. While recent diffusion-based models provided enhanced generative performance at low bitrates, they often yields limited perceptual quality and prohibitive decoding latency due to multiple denoising steps. In this paper, we propose the single-step diffusion model for image compression that delivers high perceptual quality and fast decoding at ultra-low bitrates. Our approach incorporates two key innovations: (i) Vector-Quantized Residual (VQ-Residual) training, which factorizes a structural base code and a learned residual in latent space, capturing both global geometry and high-frequency details; and (ii) rate-aware noise modulation, which tunes denoising strength to match the desired bitrate. Extensive experiments show that ours achieves comparable compression performance to state-of-the-art methods while improving decoding speed by about 50x compared to prior diffusion-based methods, greatly enhancing the practicality of generative codecs.

eess.IV↗

arXiv2022

Revealing role of Korean Physics Society with keyword co-occurrence network

Science and society inevitably interact with each other and evolve together. Studying the trend of science helps recognize leading topics significant for research and establish better policies to allocate funds efficiently. Scholarly societies such as the Korean Physics Society (KPS) also play an important role in the history of science. Figuring out the role of these scholarly societies motivate our research related with our society since societies pay attention to improve our society. Although several studies try to capture the trend of science leveraging scientific documents such as paper or patents, but these studies limited their research scope only to the academic world, neglecting the interaction with society. Here we try to understand the trend of science along with society using a public magazine named "Physics and High Technology," published by the Korean Physics Society (KPS). We build keyword co-occurrence networks for each time period and applied community detection to capture the keyword structure and tracked the structure's evolution. In the networks, a research-related cluster is consistently dominant over time, and sub-clusters of the research-related cluster divide into various fields of physics, implying specialization of the physics discipline. Also, we found that education and policy clusters appear consistently, revealing the KPS's contribution to science and society. Furthermore, we applied PageRank algorithm to selected keywords ('semiconductor', 'woman', 'evading'...) to investigate the temporal change of the importance of keywords in the network. For example, the importance of the keyword 'woman' increases as time goes by, indicating that academia also pays attention to gender issues reflecting the social movement in recent years.

physics.soc-ph↗

arXiv2021

Small-scale shear: Peeling off diffuse subhalos with gravitational waves

Subhalos at subgalactic scales ($M\lesssim 10^7 M_\odot$ or $k\gtrsim 10^3 \,{\rm Mpc}^{-1}$) are pristine test beds of dark matter (DM). However, they are too small, diffuse and dark to be visible, in any existing observations. In this paper, we develop a complete formalism for weak and strong diffractive lensing, which can be used to probe such subhalos with chirping gravitational waves (GWs). Also, we show that Navarro-Frenk-White(NFW) subhalos in this mass range can indeed be detected individually, albeit at a rate of ${\cal O}(10)$ or less per year at BBO and others limited by small merger rates and large required SNR $\gtrsim 1/γ(r_0) \sim 10^3$. It becomes possible as NFW scale radii $r_0$ are of the right size comparable to the GW Fresnel length $r_F$, and unlike all existing probes, their lensing is more sensitive to lighter subhalos. Remarkably, our formalism further reveals that the frequency dependence of weak lensing (which is actually the detectable effect) is due to shear $γ$ at $r_F$. Not only is it consistent with an approximate scaling invariance, but it also offers a new way to measure the mass profile at a successively smaller scale of chirping $r_F \propto f^{-1/2}$. Meanwhile, strong diffraction that produces a blurred Einstein ring has a universal frequency dependence, allowing only detections. These are further demonstrated through semianalytic discussions of power-law profiles. Our developments for a single lens can be generalized and will promote diffractive lensing to a more concrete and promising physics in probing DM and small-scale structures.

astro-ph.CO↗