In this paper we propose the use of seismic modeling to study the influence of pore geometry to the effective seismic velocity in carbonate rock. Velocity is both measured and modeled for various pore aspect ratio and porosity of the samples. Measurement of the velocity is performed digitally by simulating the wave propagation in the thin section photomicrographs and then measuring the corresponding effective velocity. We utilize supervised neural network to generate thin section velocity and density profile where all rock constituent effect are taken into account in the simulation. Constituent fraction needed for rock physics velocity modeling is also estimated from the generated velocity-density distribution. The modeling is undertaken by using Kuster-Toksoz model that accommodates the pore shape effect to effective elastic moduli. To be input for the model, pore shapes of the pore space are quantified by performing image analysis on the sample. The routine uses the ellipsoidal approximation of the pore shape. We observe that the measured velocity fits to those predicted by Kuster-Toksoz model, especially for sample having moderately smaller porosity and aspect ratio.