A fast CLSM undersampling image reconstruction framework with precise stage positioning for random measurements

Confocal laser scanning microscopy (CLSM) is a powerful non-destructive optical measurement system. Recently, compressive sensing (CS) is applied to the field of CLSM for high speed scan by reducing the number of sampled data required to reconstruct an accurate imaging information. However, the CS recovery algorithm employed in CLSM applications is iteration-based optimization method of which computation complexity is relatively high. In this paper, we propose a non-iteration-based deep residual convolutional neural network compressive sensing reconstruction framework (DRCNN-CSR) in end-to-end manner. Both of the computation time and the quality of reconstructed image are largely improved with this novel model. The experiment results demonstrate that our proposed method outperforms other existing reconstruction algorithm under a wide range of undersampling rates with respect to reconstruction quality comparison. In addition, CS is based on predefined random location sampling; consequently, the fast and precise positioning of scanner is required. We design the adaptive control algorithm for a piezo-driven stage to implement the CS approach in CLSM imaging; the stability of our control system design is proved by Lyapunov theorem.