Phase unwrapping aims to reconstruct the absolute phase from its values mod 2π, which is a key problem in many non-contact optical metrologies and is still a challenging issue in the presence of noise. We present a robust phase unwrapping method with two steps based on local denoising and the accumulation of residual map (ARM). In the first step, based on the local polynomial approximations and intersection of confidence intervals, we adopt a sine–cosine filtering and iteration algorithm to denoise the wrapped phase, which uses the periodicity of the trigonometric function to avoid the influence of phase jumps effectively. In the second step, the denoised wrapped phase is unwrapped by the improved ARMs method, in which we modify the edge regular term to avoid over-fitting and maintain edge information. The algorithm is tested with a set of numerical simulations under various noise levels, and the experimental results confirm its effectiveness and robustness.

中文翻译：

---

基于残差图累积和局部去噪的相位展开

相位展开旨在从其值mod2π重构绝对相位，这是许多非接触式光学计量学中的关键问题，并且在存在噪声的情况下仍然是具有挑战性的问题。我们提出了一种鲁棒的相位展开方法，该方法具有两个步骤，基于局部降噪和残差图（ARM）的累加。第一步，基于局部多项式逼近和置信区间的交集，我们采用正弦余弦滤波和迭代算法对包裹相位进行去噪，该算法使用三角函数的周期性来有效避免相位跳变的影响。在第二步中，通过改进的ARMs方法对去噪后的包裹相位进行包裹，在该方法中，我们修改了边缘正则项，以避免过度拟合并维护边缘信息。