Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Learning Continuous Spatiotemporal Implicit Neural Fields for Unsupervised Video Denoising.

Xiaowan Hu, Henan Liu, Ce Zheng

IEEE Transactions on Pattern Analysis and Machine Intelligence

|April 2, 2026

Summary

This summary is machine-generated.

Related Concept Videos

Deconvolution

Deconvolution

Deconvolution, also known as inverse filtering, is the process of extracting the impulse response from known input and output signals. This technique is vital in scenarios where the system's characteristics are unknown, and they must be inferred from the observable signals.
Deconvolution involves several mathematical techniques to derive the impulse response. One common approach is polynomial division. In this method, the input and output sequences are treated as coefficients of...

State Space Representation

State Space Representation

The frequency-domain technique, commonly used in analyzing and designing feedback control systems, is effective for linear, time-invariant systems. However, it falls short when dealing with nonlinear, time-varying, and multiple-input multiple-output systems. The time-domain or state-space approach addresses these limitations by utilizing state variables to construct simultaneous, first-order differential equations, known as state equations, for an nth-order system.
Consider an RLC circuit, a...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Large distant deletion disrupts CDKN2A enhancer and predisposes to melanoma.

medRxiv : the preprint server for health sciences·2026

Same author

Fe<sub>3</sub>O<sub>4</sub>-functionalized UiO-66 with facile magnetic recovery for high-capacity and rapid uranium removal from contaminated aqueous solutions.

Dalton transactions (Cambridge, England : 2003)·2026

Same author

Corrosion Behavior of Bubble Tubes in Glass Curing Furnaces Under the Heat-Flow Coupling Effect.

Materials (Basel, Switzerland)·2026

Same author

Disruption of CTCF binding by germline non-coding variants in <i>CDKN2B</i> suppress <i>CDKN2A</i> expression and predispose to melanoma.

medRxiv : the preprint server for health sciences·2026

Same author

A high-penetrance intergenic variant at 9p21 confers melanoma susceptibility.

Research square·2026

Same author

Accumulation and release of lead and cadmium on manganese oxides in drinking water distribution systems: Mechanism and influence factors.

Journal of hazardous materials·2026

Same journal

HardFlow: Hard-Constrained Sampling for Flow-Matching Models Via Trajectory Optimization.

IEEE transactions on pattern analysis and machine intelligence·2026

Same journal

Industrial Brain: Self-Evolving Neuro-Symbolic Autonomy with Causal Resilience for Cyber-Physical Systems.

IEEE transactions on pattern analysis and machine intelligence·2026

Same journal

Adaptive Hardness-Driven Dictionary Distillation for Incomplete Streaming View Clustering.

IEEE transactions on pattern analysis and machine intelligence·2026

Same journal

Mixture of Global and Local Experts with Diffusion Transformer for Controllable Face Generation.

IEEE transactions on pattern analysis and machine intelligence·2026

Same journal

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads.

IEEE transactions on pattern analysis and machine intelligence·2026

Same journal

Achieving Text-based Person Retrieval with Any Granularity.

IEEE transactions on pattern analysis and machine intelligence·2026

See all related articles

This study introduces a novel Spatiotemporal Implicit Neural Field (SINF) for self-supervised video denoising, overcoming limitations of existing methods. SINF effectively reduces noise and artifacts in videos with complex motion, achieving state-of-the-art results.

Area of Science:

Computer Vision
Machine Learning
Signal Processing

Background:

Existing self-supervised video denoising methods struggle with severe noise and complex motion due to discrete, grid-based representations.
Current approaches often rely on limited receptive fields or noise-sensitive optical flow, leading to artifacts and error accumulation.

Purpose of the Study:

To reformulate self-supervised video denoising using continuous spatiotemporal implicit fields.
To develop a unified model, the Spatiotemporal Implicit Neural Field (SINF), for robust video denoising.

Main Methods:

Proposed SINF leverages coordinate-based implicit neural representations for continuous spatiotemporal modeling.
Spatial domain utilizes a blind-spot implicit field for globally informed texture recovery.

Related Experiment Videos

Temporal domain employs implicit temporal embedding and a time-aware spatial graph module for continuous motion encoding and alignment.

Main Results:

SINF successfully remodels discrete video signals into a continuous spatiotemporal intensity field.
Achieved more robust pixel-wise associations compared to traditional optical flow methods.
Demonstrated state-of-the-art performance on both synthetic and real-world noisy video benchmarks.

Conclusions:

SINF offers a significant advancement in self-supervised video denoising, particularly under challenging conditions.
The continuous implicit field approach overcomes the limitations of discrete representations in existing methods.
The proposed model provides a more robust and effective solution for low-level vision and real-world imaging applications.