The performances of ghost imaging and conventional imaging in photon shot noise cases are investigated. We define an imaging signal-to-noise ratio called SNR^tran where only the object’s transmission region is used to evaluate the imaging quality and it can be applied to ghost imaging (GI) with any random pattern. Both the values SNRGItran of GI and SNRCItran of conventional imaging in photon shot noise cases are deduced from a simple statistical analysis. The analytical results, which are backed up by numerical simulations, demonstrate that the value SNRGItran is related to the ratio between the object’s transmission area A_o and the number density of photons illuminating the object plane I_o, which is similar to the theoretical results based on the first principle of GI with a Gaussian speckle field deduced by B. I. Erkmen and J. H. Shapiro [in Adv. Opt. Photonics 2, 405–450 (2010)]. In addition, we also show that the value SNRCItran will be larger than SNRGItran when A_o is beyond a threshold value.