Hybrid phase unwrapping algorithm based on a quality graph

Structured light 3D reconstruction continues to encounter significant challenges in terms of algorithmic speed and precision. Phase unwrapping, a pivotal stage in the reconstruction pipeline, directly determines the accuracy of the reconstructed model. Traditional phase unwrapping algorithms, however, exhibit limitations in suppressing noise interference and enhancing unwrapping fidelity, often resulting in suboptimal reconstruction outcomes for absolute phase. To address these issues, this study introduces an innovative quality map (BgCQuality) that integrates central curl and modulation information. Compared to existing quality maps, this novel approach, to the best of our knowledge, demonstrates heightened sensitivity to phase discontinuities and noise artifacts, providing a more robust mechanism for guiding phase unwrapping paths while mitigating adverse effects. Building upon the proposed quality map, a hybrid phase unwrapping algorithm-quality-guided flood-fill and quality-guided plane-fitting phase unwrapping-is developed. This algorithm adaptively applies quality-guided flood-fill or quality-guided plane-fitting methods to different regions of the wrapped phase map based on their quality characteristics. Experimental validation reveals that the proposed hybrid algorithm achieves a mean square error as low as 0.0008, significantly outperforming conventional flood-fill and plane-fitting methods. For high-resolution images, the algorithm executes in just 64 s, only 18 s longer than standard quality-guided methods, while reducing the accuracy error by 0.084.