Accurate and fast registration algorithm for multi-height lensless in-line on-chip holographic microscopy

Lensless in-line on-chip holographic microscope has cost-effective and compact designs that can offer sub-micron resolution over a large field-of-view. In multi-height lensless on-chip holographic microscopy, the in-line hologram series are recorded at different positions along the optical axis, thus, there inevitably exists several kinds of error sources including shifts, scales, and rotations. Therefore, digitally registering and aligning these in-line holograms is of great concern for iterative phase-retrieved recovery. Here, we proposed an accurate and fast registration method, which consists of three major steps. First, Fourier-transform-based phase correlation method is used to roughly correct the spatial translation through locating the correlation peak. Second, we proposed a morphology-based method to accurately search for the matching feature points to obtain the accurate scaling factors, where samples themselves are used as markers without any labeling. Third, the phase correlation method is used for precisely registration again. Experimental results verified the validity and effectiveness of the proposed registration algorithm. And by the comparative experiments with the existing popular registration algorithms, our proposed registration algorithm can register and align the multi-height in-line holograms with greater precision and speed.