Micro-nano Depth Information Recovery Method Based on TV Regularization

Abstract

In micro-nano computer vision, the defocus depth recovery method is used for micro-nano depth information recovery, and the Tikhonov regularization method is used to solve the problem that the objective function of the traditional defocus depth recovery method is ill-posed, resulting in low accuracy of depth information recovery. The TV regularization and L-curve method are introduced into the objective function of the traditional defocus depth recovery method. A depth information recovery algorithm based on TV regularization and L curve (L_TV algorithm) is proposed to improve the accuracy of depth information recovery. The algorithm introduces TV regularization in the objective function of the traditional defocus depth recovery method to avoid excessive punishment of the restored depth information and tends to be smooth. By introducing the L-curve method to select the appropriate regularization parameters, the depth of the recovery is avoided. The information is too smooth and retains more detail. The depth information recovery experiments of the standard 500 nm scale grid show that compared with the Tikhonov regularization method and the TSVD regularization method, the L_TV algorithm proposed in this paper can avoid excessive punishment of the recovered depth information, tend to be smooth, retain more details, and effectively improve the accuracy of deep information recovery.