The advent of deep integration and development in the domains of information technology and computational science has precipitated a paradigm shift in modern imaging technology, transitioning it from a device-driven model to one that is increasingly led by information transmission. Consequently, computational imaging has emerged as a paradigm shift, transcending the limitations of conventional technological approaches by establishing an interdisciplinary research framework that integrates multi-dimensional physical information perception and computational information interpretation. Prof. Fei Liu of Xidian University was recently invited by the editor-in-chief of Photonics Insights to co-author a paper entitled "Revolutionizing Optical Imaging: Computational imaging via deep learning". The project is being led by Prof. Chao Zuo of Nanjing University of Science and Technology, Prof. Zihan Geng of Tsinghua University, and Postdoc Yi Wei of MIT.
The review methodically and thoroughly encapsulates the synergistic strides made in the domain of computational imaging between the realms of theory and application in recent years, a period characterized by the rapid advancement of deep learning. The review focuses on the development of computational imaging, the design of computational optical systems, the interpretation of high-dimensional information in the optical field, and image processing and enhancement. It provides researchers with a valuable opportunity to understand the convergence of computational imaging and deep learning, as well as a useful reference point for future research directions.
1. Computational imaging via deep learning
Traditional optical imaging relies on enhancing physical components—higher-resolution lenses and more sensitive sensors. Yet fundamental challenges persist: the diffraction limit of light, interference from scattering media, and the absence of multidimensional information. Computational imaging revolutionizes this paradigm by redefining "imaging" not as passive light-signal recording, but as an active construction of a "light field–information–computing" closed-loop system. Through co-optimization of optical hardware and algorithms, it achieves capabilities unattainable with conventional methods. If computational imaging unlocks new frontiers for traditional optical imaging, deep learning endows it with "self-evolving" intelligence. Current breakthroughs can be distilled into three innovative dimensions: in optical design, spatial light-field modulation systems transcend the physical constraints of traditional lens arrays; in information decoding, transport equations and multidimensional light-field theory enable effective dimensionality reduction and feature extraction; in system integration, the synergy between physical constraints and data-driven optimization propels imaging technology from "hardware-dominated performance" to "information-dimensional expansion." These advancements collectively signal a paradigm shift—modern imaging science is undergoing a cognitive revolution, transitioning from passive reception to active perception.
Fig.1 Data-driven computational imaging link
2. Computational optical system design
In contrast to conventional optical imaging, computational optical system design is based on the collaborative integration of optical measurement hardware and signal processing software. This collaborative approach enables the realization of specific imaging functions while overcoming the physical limitations inherent in traditional optical systems. It also provides the adaptability and versatility required to meet diverse mission requirements. A key research focus in computational imaging is the development of more compact and stable optical imaging systems. This includes efforts to simplify designs for lighter and smaller systems, as well as leveraging photon or optical information perception to enhance imaging capabilities, as shown in Fig. 2.
Fig. 2 Applications of computational optical system design
3. Interpretation of high-dimensional information in light field
In the field of computational and optical imaging, the extraction and decoding of high-dimensional optical field information has been central to driving both technical and theoretical advancements. This chapter focuses on three core dimensions: imaging distance, imaging resolution, and information acquisition capability, as shown in Fig. 3. The limitations of distance and penetration through complex media have been addressed through techniques such as scattering imaging, non-line-of-sight imaging, polarization descattering, and laser selectivity. These methods have made it possible to observe details that were previously unobservable. Concurrently, techniques such as stacked imaging, super-resolution imaging, and holographic imaging have improved clarity at the microscopic level, enabling the precise rendering of minute structures. Additionally, polarization and spectral imaging techniques go beyond the capabilities of traditional imaging methods, allowing for the acquisition of multi-dimensional data and providing richer background information about scenes and targets. At the heart of these advancements is the interpretation of higher-order light-field information. The integration and advancement of these technologies offer a more nuanced and multifaceted understanding of the intricacies of the world, redefining how both visible and invisible environments can be explored.
Fig. 3 Applications of light-field high-dimensional information acquisition
4. Image processing and enhancement
Vision is the highest level of human perception, and images are the primary source through which humans acquire and exchange information. As such, images play an indispensable role in human perception. Computational processing, as the final link in the computational imaging chain, is characterized by high processing accuracy, good reproducibility, high flexibility, and wide applicability. It plays an increasingly critical role in obtaining high-quality images, enhancing cognitive abilities, and supporting decision-making processes. Computational imaging techniques are widely used in fields such as biomedical engineering, military and public safety, aerospace, remote sensing, and more. With advancements in computer technology and image processing, the scope of computational processing applications continues to expand. Common computational processing techniques include image fusion, image denoising, image restoration, image compression, and image enhancement. Main application areas of computational processing shown in Fig. 4.
Fig. 4 Main application areas of computational processing
5. Summary and Prospect
The driving force behind computational imaging technology is the transmission of information, with the integration of the design of the entire chain of "computational light source", "computational medium", "computational optical system", "computational detector", and "computational processing" being of particular significance. The deep integration of computational imaging and data-driven approaches at the system dimension creates a physically achievable joint optimization framework, driving the evolution of imaging systems from passive recording of "what you see is what you get" to active perception of "on-demand reconstruction". In the information dimension, the development of light field encoding theory, founded on information perception and deep learning, has enabled the efficient collection and decoupling of high-dimensional information. This has resulted in a breakthrough in various domains, including the information dimension, scale, resolution, field of view, and application scenarios. The imaging effects achieved are disruptive, with potential to achieve "higher, farther, wider, smaller, and stronger" imaging, and have the potential to greatly expand the application scope of optical imaging.
Team members unanimously recognize that the integration of computational imaging and deep learning reveals a fundamental transformation: imaging technology is evolving from "hardware-driven" to "information-driven," upgrading from "recording reality" to "reconstructing demands." When computational imaging converges with cutting-edge fields like aerospace security and biomedicine, its potential will far exceed current expectations. The domain of computational imaging extends beyond existing boundaries, with deep learning integration opening a vast horizon of possibilities. Humanity's quest for the unknown, pursuit of high-quality living, and demand for precise information will continuously propel the advancement of computational imaging, fostering its application across broader frontiers. This convergence not only redefines the limits of perception but also lays the groundwork for intelligent systems capable of adaptive, multidimensional interpretation in complex real-world scenarios.
