Dynamic foveal computational imaging based on frequency domain compressed imaging

Foveal vision, commonly observed in animal vision systems, enabling organisms to monitor a wide field of view while simultaneously capturing high-resolution images of specific local areas. This mechanism significantly enhances the efficiency of target search, recognition, and tracking under resource-constrained conditions, making it an economical visual strategy. Inspired by foveal vision, this paper proposes a dynamic foveal computational imaging method based on frequency domain compression. Through a carefully designed optical architecture, the two channels of the dual channel system are configured to share a common image plane. The system's point spread function is engineered using a genetic algorithm, allowing the mixed image captured by the system to be decomposed into individual channel images via compressive sensing theory. As a result, the system's information throughput can be doubled without increasing its bandwidth. To further enhance the efficiency of target search and tracking, the foveal channel is equipped with a scanning mirror that enables high-resolution imaging of arbitrary regions within the wide field of view channel. We present a design example of the proposed foveal computational imaging system and develop an image reconstruction algorithm based on the alternating direction method of multipliers. A proof-of-concept experiment is conducted to demonstrate the system's foveal imaging capabilities, and the experimental results verify that this method has promising potential for development in fields such as space remote sensing, target recognition, and tracking.