Abstract : A key issue in scientifc visualization is the transfer function (TF) for direct volume rendering (DVR). The TF serves as a tool for translating data values into color and opacity, to visualize the relevant structures present in the volumetric data studied. An adequate transfer function should have a non-complicated interactive strategy for new users or even experts. Furthermore, it has to achieve high-quality and not time-consuming visualization. In this paper, we propose a novel imagecentric method for the real-time generation of transfer functions. The method is based on incremental classifcation. This incremental classifcation-based approach is theoretically faster than that using batch classifcation. The method does not require users to manipulate complex widgets. We present a simple user interface adapted to the incremental learning process. Thus, this interface made it possible for the user to interact with a series of 2D images, precise the cluster, and identify some voxels. The whole volume is incrementally classifed and the rendering result is shown to the user as selected voxels are added. The TF is generated by assigning the optical properties to clusters using harmonic colors. We further introduce a novel incremental classifer, namely incremental discriminant-based support vector machine( IDSVM), that can learn through time. The IDSVM was used in the classifcation stage of the proposed image-centric method. To evaluate the IDSVM, an extensive comparison of the model with other state-of-the-art incremental and batch classifers on 12 real-world datasets and four other famous large datasets, namely MNIST-full, MNIST-test, USPS, and Fashion-MNIST, has been carried out. Using the area under curve, it has been found that the IDSVM outperforms the other classifers. Furthermore, to evaluate the proposed image-centric method, we made use of several benchmark datasets. Qualitative results and a detailed user survey demonstrate the efectiveness of the proposed method and the positive efect of the incrementality in visual and interaction time performance.