The preparation process of optical components will inevitably produce damage, with the surface undergoing scratches; the sizes of these scratches will determine the loss of quality of the instrument imaging, which can result in large errors in experiments. Since most of the information about the shapes and depths of the scratches is characterized by the phase information, acquisition of this phase information is crucial in order to detect the depth of the scratches. The classical light scattering method can only detect information about the intensity of the scratches through existing imaging devices, and cannot directly obtain the phase information for the scratches. Although the angular spectrum iteration algorithm can obtain the phase information, it has a limitation in that it is easily affected by the initial random phase. In this paper, a hybrid iterative optimized phase retrieval algorithm is proposed that combines the intensity transfer equation with the angular spectrum iterative algorithm, an approach that can improve the accuracy of the depth detection of scratches while ensuring a good convergence speed. The initial phase results from a scratched sample are first obtained by solving the intensity transport equation, and the optimized phase results and the corresponding three-dimensional distribution of the scratches are then presented using an optimization iteration module, based on amplitude optimization weighting and phase gradient descent. Finally, the required scratch depth information is obtained from the phase modulation characteristics of the scratches. The feasibility and accuracy of the method proposed in this paper are confirmed by simulation experiments and optical experiments, respectively; the root mean square error is found to be reduced from 13.14% to 4.31% and the structural similarity is improved from 84.01% to 94.46% in the scratch reconstruction experiments.