A metric calibration method for light-field three-dimensional (3-D) reconstruction based on plenoptic disk features is presented. Each of these features that relates to a single 3-D point in free physical space could be directly extracted from the raw light field. By establishing the point-to-feature correspondence, the optical parameters of unfocused plenoptic camera would be estimated through nonlinear optimization, which could then be exploited to metrically generate the 3-D point cloud of an object. For verification, all detected plenoptic disk features are reconstructed to 3-D space and then light-field reprojection is performed. In addition, to demonstrate the performance of the proposed method, comparative experiments are carried out on a commercial unfocused plenoptic camera system. Then, the acquired datasets are processed using the published method based on plenoptic disk features and our proposed calibration method, respectively. Experimental results show that the proposed calibration algorithm is more reliable and faster than the existing published one, especially for small field-of-view applications.