Efficient and Lightweight Invertible Residual Rescaling Models for High-Quality Image Reconstruction
Invertible Residual Rescaling Models (IRRM) achieve state-of-the-art performance in image rescaling tasks using a lightweight and efficient architecture.
Adapting Color Thresholds and Removing Annotations in HSV-Colored Medical Images: An Open-Source Tool for Computational Image Analysis
An open-source tool for removing letters and adapting different color thresholds in HSV-colored medical images to enable robust computational image analysis using diverse multi-center data.
Noise-Aware Illumination Interpolator for Unsupervised Low-Light Image Enhancement
A denoising-first and enhancing-later pipeline is proposed to achieve clear visibility in low-light conditions with dynamic noise, leveraging a novel noise estimation method and a learnable illumination interpolator.
Improving Texture Acutance of Digital Cameras through Hybrid Training of Denoising Networks
Hybrid training of image denoising neural networks on natural and synthetic dead leaves images can significantly improve the texture acutance metric, a standard measure of a camera's ability to preserve texture information, without impairing classic image quality metrics.
Efficient Image Fusion with State Space Model: FusionMamba
FusionMamba, an innovative method for efficient image fusion, incorporates Mamba blocks into two U-shaped networks to extract spatial and spectral features independently and hierarchically, and extends the Mamba block to accommodate dual inputs, creating a new FusionMamba block that outperforms existing fusion techniques.
Efficient Denoising of Images using Score Embedding in Score-based Diffusion Models
The core message of this paper is to improve the training efficiency of score-based diffusion models for image denoising by solving the log-density Fokker-Planck equation numerically to compute the score before training, and embedding the pre-computed score into the image to encourage faster training.
Enhancing Image Harmonization with Latent Diffusion Models
A method is proposed to enable pre-trained latent diffusion models to achieve state-of-the-art results on the image harmonization task by addressing the image distortion issue caused by the VAE compression.
Mansformer: An Efficient Transformer with Mixed Attention for Image Deblurring and Beyond
The proposed Mansformer combines multiple self-attentions, gate, and multi-layer perceptions (MLPs) to efficiently explore and employ more possibilities of self-attention for image deblurring and other restoration tasks.
Latency-aware Image Processing Transformer (LIPT): Achieving Real-time Inference with State-of-the-art Performance
LIPT is a novel latency-aware image processing transformer architecture that achieves real-time inference with state-of-the-art performance on multiple image processing tasks.
Efficient Discretization and Stability Analysis of Anisotropic Diffusion Stencils for Image Processing
The authors derive a class of finite difference discretisations on a 3×3 stencil for anisotropic diffusion processes, which split the 2-D anisotropic diffusion into four 1-D diffusions. This stencil class has one free parameter and covers a wide range of existing discretisations. The authors also establish a bound on the spectral norm of the matrix corresponding to the stencil, which allows deriving time step size limits that guarantee stability of an explicit scheme in the Euclidean norm. Furthermore, the directional splitting enables a natural translation of the explicit anisotropic diffusion scheme into a ResNet block, enabling simple and efficient parallel implementations on GPUs.
© 2024 by Linnk AI