Related Experiment Video
Updated: May 24, 2026
07:05
Applying Hyperspectral Reflectance Imaging to Investigate the Palettes and the Techniques of Painters
Published on: June 18, 2021
SGCM-Net: structure-guided mural inpainting with state space model
Yiyin Qiu1, Jianjun Chen2, Yan Fan3
1School of Computer Science, Jiangsu University of Science and Technology, Zhenjiang, 212100, China.
Scientific Reports
|May 22, 2026
Summary
This study introduces a novel Convolutional Neural Network (CNN)-Mamba hybrid for digital image inpainting, improving mural restoration by effectively handling complex textures and ensuring global consistency in degraded artworks.
Area of Science:
- Digital image restoration
- Artificial intelligence in art conservation
- Computer vision for cultural heritage
Background:
- Degraded ancient murals present restoration challenges due to intricate textures and curves.
- Traditional restoration methods are labor-intensive, while existing digital image inpainting techniques, particularly Convolutional Neural Network (CNN)-based ones, struggle with global consistency on complex structures.
- Limited receptive fields in CNNs hinder their ability to capture long-range dependencies crucial for realistic inpainting.
Purpose of the Study:
- To propose a novel CNN-Mamba hybrid inpainting architecture for restoring degraded ancient murals.
- To address the limitations of existing methods in handling complex textures and maintaining global consistency.
- To improve the accuracy and visual coherence of digital image inpainting for cultural heritage applications.
Main Methods:
- A two-stage task decomposition paradigm is employed for mural inpainting.
- Structure-Guided Fusion Blocks (SGFBs) adaptively integrate structural information from edge inpainting across multiple scales.
- Multi-Way Mamba Process Blocks (MMPBs), leveraging State Space Models (SSMs), are integrated into the network bottleneck to capture global dependencies with linear complexity.
Main Results:
- The proposed CNN-Mamba hybrid architecture demonstrates effective restoration of global styles in ancient murals.
- The method successfully fills in coherent and contextually appropriate details within degraded regions.
- Evaluations on mural and landscape painting datasets show competitive performance against established inpainting techniques.
Conclusions:
- The CNN-Mamba hybrid approach offers a significant advancement in digital image inpainting for complex artworks like ancient murals.
- The integration of Mamba-based State Space Models enhances the capture of global context, leading to improved restoration quality.
- This method provides a promising solution for preserving and restoring cultural heritage through advanced AI techniques.
Related Concept Videos
01:27State 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...
Consider an RLC circuit, a...
01:21State Space to Transfer Function
The conversion of state-space representation to a transfer function is a fundamental process in system analysis. It provides a method for transitioning from a time-domain description to a frequency-domain representation, which is crucial for simplifying the analysis and design of control systems.
The transformation process begins with the state-space representation, characterized by the state equation and the output equation. These equations are typically represented as:
The transformation process begins with the state-space representation, characterized by the state equation and the output equation. These equations are typically represented as: