In 2011, a Qihe human skull with the mandible, dating to about 10000 years ago, was found in Zhangping, Fujian Province. The skull exhibits mixed physical characteristics of both Neolithic Southern and Neolithic Northern populations, together with some primitive traits of Late Pleistocene human fossils; therefore, it has become one of the most important materials to investigate Neolithic southern human physical characteristics and migration in the Late Pleistocene to Holocene transition. There are significant morphological differences between the Qihe and recent humans; for example, the Qihe skull has a long cranial vault, a larger cranial cavity, a higher and narrower face, a wider and shorter nose bone, and lower orbits. In this paper, we propose a novel method to generate the 3D facial appearance of the Qihe human skull based on the craniofacial morphology relationship of recent human and anthropologist’s findings. Our method consisted of three steps. (1) High-resolution computed tomography (CT) was used to acquire series of images, and commercial software Mimics 16.0 was used to create the digital model of the Qihe skull. Because Qihe skull had been divided into the maxilla and mandible, we interactively reassembled these two parts and transformed the whole skull into a standardized Frankfurt Horizontal coordinate system. After exterior point clouds of Qihe skull were extracted, the Iterative Closest Point (ICP) algorithm and generic thin-plate spline (TPS) non-rigid algorithm were used to alignment the Qihe skull with one average skull from the recent human database. The dense corresponding point clouds of the Qihe skull were used to perform the craniofacial reconstruction. (2) We then applied principal component analysis (PCA) to reduce data dimensions and to calculate parameters of the skull and face of each sample in the recent human database. Least square regression was used to quantitatively represent the craniofacial morphology of the recent human. Then, the parameters of Qihe skull were input to predict three-dimensional facial appearance. Facial generation face of this step was found to be consistent and objective. (3) Because craniofacial morphology of the Qihe and recent human was similar but inconsistent, we developed an easy-to-use interactive tool to edit the facial appearance based on statistical shape model and an anthropologist’s findings. PCA was again used to construct a facial statistical shape model based on the dense corresponding point clouds. Variations in these point clouds, which were captured in every principal component (PC), were analyzed to observe shape variability. We changed the coefficients of corresponding PCs to produce multiple facial appearances and choose the best three-dimensional facial appearance. Finally, the geometric shape of the neurocranium and facial morphological characters of the Neolithic human were vividly shown in a handmade drawing. In comparison to manual craniofacial reconstruction, our method is faster, easier, and more efficient for producing facial appearance of Qihe skull via craniofacial morphology and anthropologist’s finds. Using the proposed method, the facial appearance of Qihe skull is displayed. Furthermore, our method provides efficient technical support and references for craniofacial reconstruction of ancient human skulls, which may further be applicable for other fossils and skeletons.