Multi-Scale Interactive Network with Color Attention for Low-Light Image Enhancement
Haoxiang Lu, Changna Qian, Ziming Wang, Zhenbing Liu · 2025
Read Multi-Scale Interactive Network with Color Attention for Low-Light Image Enhancement by Haoxiang Lu, Changna Qian, Ziming Wang, Zhenbing Liu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12787739.txt.
Description
Haoxiang Lu and three co-authors present MSINet, in a 2025 paper, a network that brightens low-light photos while keeping their colors true. Most earlier tools trade one gain for the other: they lift light but skew color. Photos also hold different information at fine and broad scales, which a single path reads poorly.
Their network runs two branches at once. A convolutional branch reads local detail with stacked attention blocks. A Transformer branch reads the whole scene across scales. A fusion step lets each pair of blocks exchange what they found. A third branch, built on self-attention, learns the color distribution of the dim input, which guards the hue of the output image. Wide tests set the network past the best earlier tools on both light enhancement and color accuracy. The authors aim it at vision tasks that must work in the dark.
Catalog record
| CREATOR | Haoxiang Lu, Changna Qian, Ziming Wang, Zhenbing Liu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |