Single Image Reflection Removal Based on Reflection Region
Lan Zhou, Qin Lu, Zheng Zhang, Bochuan Zheng · 2022
Images taken through transparent glass usually contain reflections, which will produce visual noise and reduce image quality. Aiming at the problem of reflection affecting image quality, we propose a method based on reflection region to remove reflections. Firstly, according to the feature of small gradient in the reflection region, the initial reflection region marker map is obtained by gradient image; Then, the more accurate reflection region marker map is obtained by using both the background region marker map and the maximum chromaticity difference binary map; Finally, according to the reflection region marker map to suppress the reflection in the reflection region. The comparison experiments with objective and subjective evaluations show that our method is better at removing reflection and retaining background details than several state-of-the-art methods.