A method for bending and strain analysis of micro-cantilever with initialization optimization by digital image correlation was proposed. Feature points were detected and matched in the scale space based on the image context obtained under microscope. First, the transformation parameters were calculated by digital image correlation with high order in the feature points. Then, the initial transformation parameters of every pixel of the cantilever were obtained by the interpolation in twodimension which could be used as the initial value for the iterative spatial domain cross correlation algorithm (NewtonRaphson method). Finally, the refined displacement and strain information at some special grid  were calculated. While the algorithm parameters were optimized in the experiment of the simulated bended images, the performance of the proposed method was verified through the strain comparison with the finite element method.