Specular highlight suppression based on full polarization imaging

Specular highlights severely hinder the effective extraction of target features in industrial inspection. Existing methods for specular highlight suppression struggle to surfaces with high reflective characteristics. Considering that the specular reflection from metal surfaces contains a substantial amount of circularly polarized light under linear polarized lighting, this paper proposes a full polarization imaging-based method for specular highlight suppression. First, a full polarization image dataset is constructed using the quarter-waveplate rotation method, and an analytical model for full polarization feature parameters is derived. Second, an adaptive specular highlight detection method based on clusters of polarization features is proposed, and a region expansion rule is introduced to include the transition area between specular and diffuse reflections. Finally, a compensation value calculation method based on edge curve integration is used to eliminate mismatches during the image stitching process, ultimately producing specular highlight suppression images. Our method was evaluated on both our self-built image dataset FPID and publicly available datasets. Experimental results demonstrate the superiority of our method in both visual effects and metrics.