Related Experiment Video
Updated: Aug 22, 2025
04:47
Multimodal Optical Imaging Platform for Studying Cellular Metabolism
Published on: June 6, 2025
578
4-Band Multispectral Images Demosaicking Combining LMMSE and Adaptive Kernel Regression Methods
Norbert Hounsou1, Amadou T Sanda Mahama1, Pierre Gouton2
1Institute of Mathematics and Physical Sciences, University of Abomey-Calavi, Porto-Novo BP 613, Benin.
Journal of Imaging
|November 10, 2022
Summary
This study introduces a novel multispectral demosaicking algorithm for improved image reconstruction. The method enhances visual quality and quantitative metrics like PSNR, SSIM, and RMSE.
Area of Science:
- Image Processing
- Computer Vision
- Optics
Background:
- Multispectral imaging systems are expanding with diverse demosaicking algorithms.
- Optimal multispectral demosaicking is crucial for accurate raw image reconstruction.
Purpose of the Study:
- To present a novel four-band multispectral filter array (MSFA) and demosaicking algorithm.
- To improve image reconstruction accuracy and reduce errors from single-sensor raw images.
Main Methods:
- Developed a four-band MSFA with a dominant blue band.
- Combined Linear Minimum Mean Square Error (LMMSE) for blue band estimation.
- Utilized directional gradient methods and adaptive kernel regression for spectral band updates.
Main Results:
- The proposed method demonstrated superior visual and quantitative performance.
- Achieved higher Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and lower Root Mean Square Error (RMSE).
Conclusions:
- The novel demosaicking algorithm effectively reconstructs multispectral images.
- Outperforms existing methods in accuracy and artifact reduction.
