Measurement and 3-D Image Reconstruction Using a Novel Multiview Optical Instrument and Semiadaptive Scale-Invariant Feature Transform

Abstract

Many studies rely heavily on the 3-D reconstruction of optical microscope (OM) images; however, conventional approaches such as Z-stacking are limited to a single viewing angle, which inevitably results in visual blind spots. Moreover, OM systems often struggle to reliably match keypoints when processing samples with weak texture features, thereby degrading the overall performance of 3-D reconstruction. Consequently, multiview 3-D reconstruction remains highly challenging. To address this issue, this article proposes a multiview OM system co-designed with both hardware innovations and an advanced 3-D reconstruction algorithm to reduce visual blind spots and improve reconstruction performance for weak-texture samples. By continuously capturing and fusing multiview images at different focal distances, the proposed system enhances local contrast while preserving structural edges. Prior to keypoint extraction, the degree of smoothing is adaptively adjusted according to local texture characteristics to retain meaningful information in low-texture regions. The proposed multiview OM system integrates the developed 3-D reconstruction algorithm and is experimentally evaluated in terms of feature detection, matching performance, and 3-D reconstruction quality.