The sharpness evaluation of the digital holographic reconstructed image is the key problem of holographic auto-focusing. This paper proposes a sharpness evaluation method for digital holographic auto-focusing of micro-diameter milling tools, and the focusing process is divided into coarse focusing and fine focusing. In coarse focusing, the search step of the reconstructed distance calculation is in millimeter level, and the edge of the reconstructed image will have obvious oscillation, which cannot be accurately judged by the traditional sharpness evaluation function, but the amplitude of the wavelet transform high frequency coefficient of the most clearly focused image in the focusing window is the largest, and the sharpness evaluation of the reconstructed image can be accurately performed by using this characteristic. When fine focusing, the search step of the calculation of the reconduction distance is adjusted to micron level, and by obtaining the evaluation results of different evaluation functions, it is found that different evaluation functions can be evaluated accurately.