SINGLE IMAGE REFLECTION REMOVAL WITH SEGMENTATION

Abstract
Removal of reflection is of high importance to reclaim the original background image. Several attempts have been made to separate reflection from background. A number of approaches are based on assuming certain conditions about the reflective material (glass) and type of reflection. Humans can separate familiar objects easily due to the understanding of the objects in scene, same analogy is applied here. In this paper, additional information of segmentation map is utilized rather than using a single reflection image as input. Estimated segmentation map corresponds to the composite image. Our aim is to investigate the efficacy of segmentation map in reflection removal approaches. Proposed method performs adequately on real-world images and suppresses the reflection components in background effectively.

Authors
Rashmi Chaurasiya1, Dinesh Ganotra2
Indira Gandhi Delhi Technical University for Women, India1,2

Keywords
Reflection Removal, Deep Learning, Semantic Guidance
Published By :
ICTACT
Published In :
ICTACT Journal on Image and Video Processing
( Volume: 12 , Issue: 4 )
Date of Publication :
May 2022
DOI :

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.