When the reconstructed image is out of focus, oscillation will appear at the edge of objects, because digital holography is a coherent imaging system. Therefore, it is inaccurate to perform auto-focusing by the traditional evaluation function of definition in digital holography. It can be found through wavelet analysis that the amplitude of high-frequency coefficients out of focus is smaller than that in focus. The evaluation function of definition based on Laplacian operator and wavelet transform is improved. The maximum amplitude of high-frequency coefficients in the focal windows, instead of the sum of high-frequency coefficients, is used to evaluate the definition of the reconstructed image. In order to get a clear image that contains all relevant objects in the three-dimensional (3D) space of digital holography, a fusion algorithm is proposed. The auto-focusing and reconstructed image fusion experiments for simulated digital in-line holography and an ant detection experiment are completed. Experimental results show that the improved evaluation function of definition can accurately achieve auto-focusing for digital holography, and a series of reconstructed image can be fused into a clear image by the proposed fusion algorithm.