Optimization of Depth from Defocus Based on Iterative Shrinkage Thresholding Algorithm

In solving the dynamic optimization of depth from defocus with the iterative shrinkage threshold algorithm (ISTA), the fixed iteration step decelerated the convergence efficiency of the algorithm, which led to inaccuracy of reconstructed microscopic 3D shape. Aiming at the above problems, an optimization of ISTA algorithm based on gradient estimation of acceleration operator and secant linear search (FL-ISTA) was proposed. Firstly, the acceleration operator, which consists of the linear combination of the current and previous points, was introduced to estimate the gradient and update the iteration point. Secondly, the secant linear search was used to dynamically determine the optimal iteration step, which accelerated the convergence rate of solving the dynamic optimization of depth from defocus. Experimental results of standard 500 nm grid show that compared with ISTA, FISTA and MFISTA algorithms, the efficiency of FL-ISTA algorithm was great improved and the depth from defocus decreased by 10 percentage points, which close to the scale of 500 nm grid. The experimental results indicate that FL-ISTA algorithm can effectively improve the convergence rate of solving dynamic optimization of depth from defocus and the accuracy of the reconstructed microscopic 3D shape.