Purpose: In most photoacoustic (PA) tomographic reconstructions, variations in speed-of-sound (SOS) of the subject are neglected under the assumption of acoustic homogeneity. Biological tissue with spatially heterogeneous SOS cannot be accurately reconstructed under this assumption. The authors present experimental and image reconstruction methods with which 2D SOS distributions can be accurately acquired and reconstructed, and with which the SOS map can be used subsequently to reconstruct highly accurate PA tomograms. Methods: The authors begin with a 2D iterative reconstruction approach in an ultrasound transmission tomography setting, which uses ray refracted paths instead of straight ray paths to recover accurate SOS images of the subject. Subsequently, they use the SOS distribution in a new 2D iterative PA reconstruction approach, where refraction of rays originating from PA sources is accounted for in accurately retrieving the distribution of these sources. Both the SOS reconstruction and SOS-compensated PA reconstruction methods utilize the Eikonal equation to model acoustic wavefront propagation. The equation is solved using a high accuracy fast marching method. Results: The authors validated the new reconstruction algorithms using numerical phantoms. For experiments they utilized the recently introduced PER-PACT method which can be used to simultaneously acquire SOS and PA data from subjects. Conclusions: It is first confirmed that it is important to take SOS inhomogeneities into account in high resolution PA tomography. The iterative reconstruction algorithms, that model acoustic refractive effects, in reconstructing SOS distributions, and subsequently using these distributions to correct PA tomograms, yield artifact-free highly accurate images. The approach of using the hybrid measurement method and the new reconstruction algorithms is successful in substantially improving the quality of PA images with a minimization of blurring and artifacts.