At present, the porosity of carbon fiber reinforced plastic (CFRP) is usually determined by measuring ultrasonic attenuation coefficient. The available detection models mainly focus on the ultrasonic scattering mechanism from regular voids, and pay little attention to the real random voids. However, some researches have demonstrated that the interaction between natural voids and ultrasonic beam is much more complicated than that of regular voids. Due to random and complex morphology of actual voids, is it is extremely difficult to precisely describe the voids as well as their effects on ultrasonic scattering. In this paper, based on image processing technology and statistical methods, a real morphology void model (RMVM) was established for CFRP laminates. A series of photomicrograph with the porosity from 0.58% to 3.49% were simulated to investigate the relationship between porosity P and ultrasonic attenuation coefficient α. Simulated results showed that the P-α scatter diagram presented a banded distribution, and the fluctuation range of attenuation coefficient enlarged with the increase of porosity. The non-unique corresponding relationship between P and α was further verified by experiments. It is concluded that the complex void morphology affects the relationship between porosity and ultrasonic attenuation coefficient.