Specular highlight removal of light field based on dichromatic reflection and total variation optimizations. (April 2022)

• Record Type: Journal Article
• Authors: Feng, Wei; Li, Xiuhua; Cheng, Xionghao; Wang, Henghui; Xiong, Zhi; Zhai, Zhongsheng
• Is Part Of: Optics and lasers in engineering. Volume 151(2022)
• Publication Date: 2022-04
• Subjects: Light field camera -- Highlight removal -- Dichromatic reflection model -- Total variation -- Depth reconstruction; Lasers in engineering -- Periodicals; Optical measurements -- Periodicals; Optics -- Periodicals; Lasers en ingénierie -- Périodiques; Mesures optiques -- Périodiques; Optique -- Périodiques; 621.36605
• DOI: 10.1016/j.optlaseng.2021.106939
• Languages: English
• ISSNs: 0143-8166

Abstract

Highlights: An approach that combines dichromatic reflection model (DRM) with light field based on nonlinear least square method is proposed to separate diffuse reflectance information in highlight region. A total variation (TV) optimization model is designed to remove the texture aliasing by regarding occlusion in texture area as noise to obtain the diffuse image without highlights. A method named SRE that combines specular residual ratio with image entropy is built to quantitatively evaluate the quality of highlights removal. The experimental process and results analysis are presented in detail. Abstract: In this paper, a novel and effective method based on dichromatic reflection model (DRM) and total variation (TV) optimization with light field camera is proposed to remove specular highlight. The unique feature of proposed method is that both diffuse components and specular components can be obtained by analyzing the tristimulus integration of DRM, which uses the characteristics of multiple views of the light field camera. In addition, the TV optimization model based on local noise window is established to eliminate texture aliasing through regarding the occlusion of local window as noises. After that, a strategy of specular-free-ratio and image entropy is designed to quantitatively evaluate the quality of highlight removal. Finally, extensive experiments including visual imaging and depth reconstruction for comparison with the state-of-the-art methods show that the proposed method can effectively restore the highlight information, which will be useful for 3D reconstruction of highly reflective surfaces. Generally, our method can fully detect and process all highlights with stronger robustness, and outstanding resilience for challenging or noising scenes.