World PulseNow
Newsletter

4KDehazeFlow: Ultra-High-Definition Image Dehazing via Flow Matching

arXiv — cs.CV•Thursday, November 13, 2025 at 5:00:00 AM
PositiveArtificial Intelligence
4KDehazeFlow represents a significant leap in Ultra-High-Definition image dehazing, tackling challenges faced by previous methods, such as limited adaptability and high computational demands. This innovative approach employs Flow Matching and a haze-aware vector field to optimize the dehazing process through a continuous vector field flow. Its compatibility with various deep learning networks and the introduction of a learnable 3D lookup table streamline the haze transformation process. The method's performance is validated through extensive experiments, demonstrating a notable improvement over seven leading techniques, with a PSNR increase of 2dB and superior results in dense haze scenarios. This development is vital for enhancing image quality in numerous applications, from photography to surveillance, where clarity and detail are paramount.
— via World Pulse Now AI Editorial System
Read Original

Was this article worth reading? Share it

Recommended Readings
Parameter-Efficient MoE LoRA for Few-Shot Multi-Style Editing
arXiv — cs.CV2 days ago
Parameter-Efficient MoE LoRA for Few-Shot Multi-Style Editing
PositiveArtificial Intelligence
The paper titled 'Parameter-Efficient MoE LoRA for Few-Shot Multi-Style Editing' addresses the challenges faced by general image editing models when adapting to new styles. It proposes a novel few-shot style editing framework and introduces a benchmark dataset comprising five distinct styles. The framework utilizes a parameter-efficient multi-style Mixture-of-Experts Low-Rank Adaptation (MoE LoRA) that employs style-specific and style-shared routing mechanisms to fine-tune multiple styles effectively. This approach aims to enhance the performance of image editing models with minimal data.
Read full article
via arXiv — cs.CV
Flow matching-based generative models for MIMO channel estimation
arXiv — cs.LG2 days ago
Flow matching-based generative models for MIMO channel estimation
PositiveArtificial Intelligence
The article presents a novel flow matching (FM)-based generative model for multiple-input multiple-output (MIMO) channel estimation. This approach addresses the slow sampling speed challenge associated with diffusion model (DM)-based schemes by formulating the channel estimation problem within the FM framework. The proposed method shows potential for superior channel estimation accuracy and significantly reduced sampling overhead compared to existing DM-based methods.
Read full article
via arXiv — cs.LG