Superresolution reconstruction based on complementary composite structural illumination

Structured illumination moves high-frequency information with more details into the frequency band that the optical system can pass through, so as to achieve high-quality image super-resolution (SR) reconstruction. Although structured illumination microscopy (SIM) using this technique are already mature for optical SR imaging, continuous illumination through unidirectional multi-frame sinusoidal structured patterns greatly reduces temporal resolution and poses challenges for robust imaging in low-light and short-exposure environments. In order to introduce more high-frequency information into the original image with fewer frames, we propose an innovative illumination method using compound angle structured patterns. Only one frame of the structured pattern is loaded on a spatial light modulator (SLM), and then two images are captured by complementary modulation. The reconstruction algorithm uses end-to-end deep neural networks. In order to improve the applicability under real low-light conditions, we simulate the actual noise scenes for model training. Experimental results show that our method is robust to noise and can restore image details under low-light conditions to achieve image SR. Moreover, it effectively minimizes the exposure time of image acquisition and reduces the computing time of the network, leading to an enhancement in the efficiency of image reconstruction.