虚拟现实（Virtual Reality，VR）是一种利用头戴显示系统提供给使用者高度沉浸感的技术，在航天、军事、医疗、教育等各个领域都有着极大的应用价值。然而，人眼对于沉浸式显示的需求十分苛刻，当前的软硬件水平很难支撑现有的VR架构实现高时空分辨率的三维显示。究其原因，主要是当前的VR方案对于不同内容、不同区域进行无差别的高质量显示，导致计算显示资源的使用效率非常低下。因此如何降低系统的冗余，利用有限的算力和显示资源来提供给使用者沉浸的显示效果成为了VR系统和算法设计中最重要的问题。围绕这一问题，本论文利用人类视觉的特性，开发了高效的新型VR系统和算法，主要成果和创新点包括：（1） 针对现有VR系统计算显示资源利用率低，难以实现高时空分辨率显示的问题，本文借鉴人眼的时空间感知特性，提出了一种可以灵活分配显示资源的双模式VR系统。该系统通过合理配置反射镜、液晶快门、分束镜等光学元件使显示光路具备可调控的能力。在基于人眼时空感知模型的模式切换算法的驱动下，该系统能够根据用户的注视点位置和场景的变化，灵活地切换到高速显示模式或时分复用的注视点显示模式，从而高效地利用软硬件资源，实现感知上的高时空分辨率显示。（2）针对VR系统难以提供正确深度信息，造成使用者眩晕的问题，本文借鉴人眼深度感知的空间特性，提出了一种注视点变焦的VR显示方案。该方案设计了基于液体透镜的中继成像镜头，该镜头会根据用户的注视点位置，将微显示器成像于相应的深度，实现注视点区域的高清变焦显示。同时还设计了一种光学耦合棱镜，让该棱镜与前述中继成像镜头协同工作，构造出紧凑的光学系统，从而还可以在外围区域实现大视场的显示。两路显示在自适应渲染算法的调度下，能在整个视野范围内模拟离焦模糊效果，高效地解决VR显示深度的问题。（3） 在运用VR系统进行生命科学数据可视化分析的应用中，针对目前细胞超分辨成像数据采集效率低、难以实时成像的问题，提出了一种高效无伪影的结构光超分辨重构方案。该方案修正了传统成像模型中偏振的影响，并设计了统一使用多方向结构光数据的迭代重构算法。该算法能够灵活使用4-9张原始图像实现不同质量无伪影的超分辨重构，因此可以在VR观测的实时三维成像应用中，根据使用者的注视点位置灵活地调整重构的时空间分辨率，实现高效地数据采集。

Virtual Reality (VR) is a technology that provides users with a highly immersive experience using head-mounted display systems. It has immense application value in various fields such as aerospace, military, healthcare, and education. However, the demands of human eyes for immersive displays are stringent, and the current hardware and software are difficult to support the high spatio-temporal resolution three-dimensional display of existing VR architectures. The main reason for this lies in the indiscriminate high-quality display of different content and regions in current VR solutions, leading to extremely low efficiency in the utilization of computational and display resources. Therefore, effectively reducing redundancy in the system and providing users with an immersive display experience using limited computing power and display resources has become the most important problem in VR system and algorithm design. To address this issue, this thesis leverages human visual characteristics to develop efficient new VR systems and algorithms. The main research contents and innovative work of this thesis are as follows:（1） To address the issue of low utilization of computational display resources in existing VR systems and the difficulty in achieving high spatio-temporal resolution displays, this study proposes a dual-mode VR system inspired by the spatio-temporal perception characteristics of the human eye. This system enables flexible allocation of display resources by configuring optical components such as mirrors, liquid crystal shutters, and beam splitters, providing the display optics with adjustable capabilities. Coupled with a mode-switching algorithm based on the spatio-temporal visual acuity model of the human eye, the system can adaptively switch between high-speed display mode and time-multiplexing foveated display mode based on the user‘s gaze point position and scene changes. This enables efficient utilization of software and hardware resources to achieve perceptually high spatio-temporal resolution displays.（2） To addressing the issue that existing VR system is difficult to provide accurate depth information, which leads to dizziness of users, this paper draws inspiration from the spatial characteristics of human eye depth perception and proposes a foveated varifocal VR display scheme. The proposed solution involves designing a relay imaging lens based on liquid lenses, which images the micro-display at the corresponding depth according to the user‘s focal point position, enabling high-definition varifocal display in the foveal region. Additionally, an optical coupling prism is designed to work in conjunction with the aforementioned relay imaging lens, creating a compact optical system that can achieve wide-field display in peripheral areas. With the coordination of two display paths under an adaptive rendering algorithm, this setup can simulate defocus blur effects across the entire field of view, effectively addressing the depth perception issues in VR displays.（3） In the application of using VR systems for three-dimensional visualization and analysis of life science data, a high-efficiency and artifact-free structured illumination microscopy (SIM) reconstruction scheme is proposed to address the problems of low efficiency and difficulty in real-time analysis of cell super-resolution (SR) multidimensional imaging data. This scheme utilizes improved SIM SR imaging model that considers the polarization-dependent absorption efficiency and designs an iterative reconstruction algorithm that can jointly exploit the raw SIM images of all illumination orientations. The reconstruction algorithm is very flexible and can be used to reconstruct using 4-9 raw SIM images. Therefore, in real-time three-dimensional imaging applications of VR observation, the spatial and temporal resolution of reconstruction can be flexibly adjusted according to the user‘s gaze point position, achieving efficient SR data collection.