Realistic 3D microstructure models of wood fiber networks (WFNs, e.g., paper, molded fibers, hot-pressed fibers, etc.) are of interest for numerical modeling of mechanical, optical, and other physical properties. One challenge is to numerically describe 3D high-density WFN (HD-WFN) models with complex fiber shapes without fiber overlapping. An efficient method is proposed to generate 3D HD-WFN microstructures with porosities as low as 21.5% while without fiber overlapping. The HD-WFN microstructures are obtained by compressing an initially sparse structure using a geometrically designed 3D displacement field. The sparse structure can be optimized to obtain a very low-porosity HD-WFN by reducing the local fiber clustering. The method can generate HD-WFNs with designated shapes (e.g., molded shapes) and varied structural parameters, including fiber width, orientation, location, and material thickness. Each interfiber bond and local material axis in the HD-WFN models can be determined for numerical simulation of properties. The optical scattering of transparent HD-WFN models (polymer matrix composites, transparent paper) is numerically studied using ray tracing methods. The open-source code is available on GitHub, and it can be used to obtain a large dataset for deep learning modeling.