Design method of a lens with an extended depth of field based on multimodal deep optics

End-to-end optical design has become an emerging paradigm in computational imaging, which seeks jointly optimal solutions by co-optimizing the optical design and computational algorithm. However, most existing works focus on single-mode optical configurations and exhibit weakness in handling complex parameters when handling multimodal optical tasks. Here, we propose an end-to-end polymorphic lens design method based on multimodal deep optics, which establishes a differentiable multi-configurations model for the optical system and evaluates the integrated loss function. We demonstrate the effectiveness of this approach through the design of an optical system with extended depth of field (EDOF) based on a liquid lens with a working distance from infinity to 3 cm. We believe this method holds a broad application potential in zoom lens design.