Past research has demonstrated that a static, three-dimensional (3D) object scene can be directly recorded as a complex digital hologram. However, numerical reconstruction of the object scene, which may comprise multiple sections located at unknown distances from the hologram, is a complicated and computation-intensive process. To the best of our knowledge, we propose, for the first time, a low complexity method that is capable of reconstructing a complex hologram, such that sections at different depths in the 3D object scene can be automatically reconstructed at the correct focal distances and merged into a single image for an extended depth of field. We demonstrate an order of magnitude increase of the depth of field for binary objects. With the use of a graphical processing unit, the reconstruction of a $512\times 512$ complex hologram can be accomplished in about 100 ms, equivalent to around 10 frames per second.