Significance  Since Gauss’s Day, the design and development of imaging systems have been continuously dedicated to the iterative improvement and optimization of lenses, aiming to collect light emitted in various directions from a point on the object plane and converge it as perfectly as possible onto a point on the image plane. However, imaging sensors can only capture and record the spatial position information of the light field, losing the angular information, which causes them to completely lose the ability to perceive perspective transformation and depth of a three-dimensional scene. To compensate for this deficiency, computational light field imaging technology was born, which can record the complete distribution of the radiance, and jointly recording spatial position and angle information, breaking through the limitations of classical imaging. It is gradually being applied to fields such as life sciences, national defense security, virtual reality/augmented reality, and environmental monitoring, with important academic research value and broad application potential. However, light field imaging technology is still jointly constrained by digital imaging devices and image sensors. The limited spatial bandwidth product (SBP) of the imaging system makes light field imaging often have to make trade-offs between spatial resolution and angular resolution in practical applications, making it difficult to achieve the high spatial resolution of traditional imaging technology. Since the birth of light field imaging technique, how to endow it with higher degrees of freedom, that is, to maintain high-resolution imaging while improving temporal resolution and angular resolution, in order to achieve clearer and more stereoscopic imaging performance, is a key problem that light field imaging technology urgently needs to solve, and has always been a hot topic in this field.

Progress  We first reviews the development history of light field imaging technology, and elaborates in detail on the basic concepts of the seven-dimensional full light field function and the simplified four-dimensional light field. Subsequently, the paper delves into the latest research progress of light field imaging technology in enhancing temporal, spatial, and angular resolution, including the application of microlens arrays, phase scattering plates, heterodyne coding, camera arrays, and other methods in the recording of high-speed dynamic three-dimensional scenes; the achievements of techniques such as transfer function deconvolution, prior information constraints, multi-frame scanning, aperture coding, confocal, and hybrid high/low light field imaging in improving spatial resolution; and methods for enhancing angular resolution through depth constraints, sparse prior constraints, and other approaches. In addition, the article also looks forward to the future development directions of light field imaging technology, including the control of high-dimensional coherence in light fields, the integration of artificial intelligence with light field imaging technology, the development of miniaturized and portable light field imaging devices, new imaging mechanisms, and the prospects for the combination of light field imaging and light field display technology, as well as the potential of light field imaging in special fields. Finally, the paper emphasizes the challenges of light field imaging technology in achieving optimal imaging performance, especially the importance of finding an efficient balance among the three key dimensions of temporal, spatial, and angular resolution. With the continuous progress and innovation of technology, light field imaging technology is expected to further break through the limitations of traditional optical imaging, inject new momentum into the advancement of imaging technology, and open up new application prospects in various fields such as life sciences, remote sensing, computational photography, and spectral imaging.

Conclusions and Prospects  This review delves into the current state of development and challenges faced by light field imaging technology. Light field imaging still utilizes existing two-dimensional sensor devices, which currently possess only spatial resolution, thus necessitating a trade-off between spatial and angular resolution to achieve the latter. To enhance the spatial resolution of light field imaging, one approach is to improve hardware, such as increasing the pixel resolution of sensors and arranging large-scale camera arrays. On the other hand, temporal resolution can be used to enhance spatial sampling, for instance, by employing aperture coding techniques and multi-frame scanning methods to increase spatial resolution. However, directly upgrading hardware resources can lead to increased costs, as well as potential issues with size, weight, and data transmission processing. Therefore, if one does not wish to enhance performance by physically adding hardware, it is necessary to rely on algorithmic innovation, such as optimizing imaging results using prior information and deep learning. This method is referred to as "punching above one's weight," meaning that performance is improved through algorithms without increasing physical resources. However, we also recognize that the development of light field imaging technology still faces challenges on multiple fronts. To achieve broader applications, future research needs to find a better balance between algorithmic innovation, hardware optimization, and cost-effectiveness. In summary, the future of light field imaging technology is promising, and it will continue to serve as an important tool in fields such as biomedical imaging, materials science, and industrial inspection. Through interdisciplinary collaboration and innovative thinking, we believe that more efficient and accurate imaging technologies can be realized, opening up new horizons for scientific research and practical applications.