Extracting only the visible portion of an isosurface can improve both the computation efficiency and the rendering speed. However, the visibility test overhead can be quite high for large scale data sets. In this paper, we present a view-dependent isosurface extraction algorithm utilizing occlusion query hardware to accelerate visible isosurface extraction. A spherical partition scheme is proposed to traverse the data blocks in a layered front-to-back order. Such traversal order helps our algorithm to identify the visible isosurface blocks more quickly with fewer visibility queries. Our algorithm can compute a more complete isosurface in a smaller amount of time, and thus is suitable for time-critical visualization applications. Categories and Subject Descriptors (according to ACM CCS): I.3.1 [Computer Graphics]: Hardware Architecture; I.3.7 [Computer Graphics]: Three-Dimensional Graphics and Realism I.3.7 [Computer Graphics]: visible line/surface algorithms