Automatic shadow detection and removal using image matting

• Published in: Signal processing Vol. 170; p. 107415
• Main Authors: Amin, Benish, Mohsin Riaz, M., Ghafoor, Abdul
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.05.2020
• Subjects: Automatic shadow detection; Image matting; Shadow removal; Shadow removal; Image matting; Automatic shadow detection
• ISSN: 0165-1684; 1872-7557
• DOI: 10.1016/j.sigpro.2019.107415

•Automatic method to extract and remove shadows.•Tricolor attenuation model and intensity information are used.•Trimap using a guided filter is used for rough segmentation.•Robust matting is then used to refine the shadowed region.•The shadow edges are recovered by employing inpainting technique. Extraction of shadows from a single image (also known as shadow matting) is a difficult problem and often requires user interaction. This paper presents an automatic method to extract and remove shadows from real images using the tricolor attenuation model (TAM) and intensity information. Trimap using a guided filter is used for rough segmentation of shadow region. Robust matting is then used to refine the shadowed region in the image and the shadow is removed by relighting the segmented region in RGB space. Finally, the shadow edges are recovered by employing inpainting technique near shadow boundary. Compared with existing shadow removal techniques, the proposed method accurately preserves the color and texture of the images under variety of scenarios without user intervention.