Enhanced ultrafine multimode fiber imaging based on mode modulation through singular value decomposition

Abstract

The utilization of multimode fibers (MMFs) displays significant potential for advancing the miniaturization of optical endoscopes. However, the imaging quality is constrained by the physical conditions of MMF, which is particularly serious in small-core MMFs because of the limited mode quantity. To break this limitation and enhance the imaging ability of MMF to the maximum, we propose a mode modulation method based on the singular value decomposition (SVD) of MMF's transmission matrix (TM). Before injection into the MMF, a light beam is modulated by the singular vectors obtained by SVD. Because the singular vectors reflect the transfer characteristics of MMF and selectively excite the modes of different orders, the optimal distribution of the excited modes in MMF can be achieved, thereby improving the imaging quality of the MMF imaging system to the greatest extent. We conducted experiments on the MMF system with 40 µm and 105 µm cores to verify this method. Deep learning is utilized for image reconstruction. The experimental results demonstrate that the properties of the output speckle pattern were customized through the selective excitation of optical modes in the MMF. By applying singular vectors for mode modulation, the imaging quality can be effectively improved across four different types of scenes. Especially in the ultrafine 40 µm core MMF, the peak signal-to-noise ratio (PSNR) can be increased by up to 6.82 dB, and the structural similarity (SSIM) can be increased by up to 0.103, indicating a qualitative performance improvement of MMF imaging in minimally invasive medicine.