This paper presents a fast computation and processing for off-axis digital holographic microscopy (DHM) based on a graphic processing unit (GPU). DHM is a highly effective imaging technique capable of reconstructing the three-dimensional (3-D) structure by retrieving the amplitude and phase information of the specimen. To obtain the 3-D profile of the specimen, numerous calculations for image reconstruction are required. For achieving enormous calculations of image reconstruction and improving the computational speed for their practical applications, we have performed the preceding computation and processing of DHM through the GPU. This method leads to effectively accelerate the computation performance with the parallel computing architecture and algorithm. Further, we perform accurate 3-D profile of the specimen by the use of filtering algorithm adjusting the contrast of reconstructed images. The proposed GPU-based DHM provides the potential for significantly speed-up of numerical calculation compared to the counterpart of the central processing unit (CPU). The high-speed cost-effective computational capability derives 3-D profile measurement in real time and on-line processing.