Multimode Fiber Imaging With Bit-Plane Encoding: Improving Fidelity and Speed

High-performance natural scene image transmission via multimode fiber (MMF) represents a substantial challenge. Conventional MMF gray imaging system relies on the visual persistence phenomenon of the digital micromirror device (DMD), causing redundant capture time and loss of bit-depth information. We introduce bit-plane encoding to extract eight speckles from gray targets at different bit-planes, thereby constructing a speckle cube to enrich the captured information. Our self-designed HySpeckNet model enhances the connections among high-dimensional features within speckle cubes. As a result, the average structural similarity index (SSIM) on the ImageNet dataset improves from 0.586 to 0.617, representing a 5.3% increase compared to traditional methods. Notably, gray image reconstruction can be approximately achieved using only two high-order speckles, resulting in a transmission speed about 11.76 times faster than traditional gray image restoration method. Our research sets the stage for the practical transmission of complex image information via MMFs.