High-quality three-dimensional visualization system for light field microscopy using a robust-depth estimation algorithm and holographic optical element-microlens array

•The high-quality 2D images and light field images of the micro-objects are simultaneously captured through the HOE-MLA using Bragg angle characteristics of the HOE recording, and dual-channel high-resolution camera.•The RDE algorithm achieved superior accuracy in estimating depth data for objects compared to existing methods, utilizing both 2D and light field images, including directional-view and refocused images. This approach aims to gather the most reliable depth information available.•Natural-view clear 3D point cloud models of objects were reconstructed by combining accurately estimated depth data with high-resolution 2D images. These models feature high-accuracy and high-quality depth information, akin to that found in 2D images. We propose a high-quality three-dimensional (3D) visualization system based on light field microscopy using a robust-depth estimation (RDE) algorithm and a holographic optical element-microlens array (HOE-MLA). According to the Bragg angle characteristics, the HOE-MLA allows the simultaneous capture of both images viewed through a dual-channel camera. This system provides high-quality 3D visualization by combining a depth map estimated from the light field with high-resolution two-dimensional (2D) images and texture information. To estimate the depth and shape of micro-objects with high accuracy, a custom-designed RDE algorithm based on reliability analysis and image matting information considers all information obtained from high-resolution 2D and directional-view images. Based on the acquired depth and color/texture information, clear, natural-view 3D visualization of the object is reconstructed, including depth information with high accuracy and high quality, similar to that of a 2D image. The superiority of the proposed system is verified experimentally, and the results indicate that it can be used to easily acquire and display high-quality 3D information of microscopic objects with simple optical structure.