There remains an urgent need to develop effective photoacoustic computed tomography (PACT) image recon- struction methods for use with acoustically inhomogeneous media. Transcranial PACT brain imaging is an im- portant example of an emerging imaging application that would benefit greatly from this. Existing approaches to PACT image reconstruction in acoustically heterogeneous media are limited to weakly varying media, are computationally burdensome, and/or make impractical assumptions regarding the measurement geometry. In this work, we develop and investigate a full-wave approach to iterative image reconstruction in PACT for media possessing inhomogeneous speed-of-sound and mass density distributions. A key contribution of the work is the formulation of a procedure to implement a matched discrete forward and backprojection operator pair, which facilitates the application of a wide range of modern iterative image reconstruction algorithms. This presents the opportunity to employ application-specific regularization methods to mitigate image artifacts due to mea- surement data incompleteness and noise. Our results establish that the proposed image reconstruction method can effectively compensate for acoustic aberration and reduces artifacts in the reconstructed image.