Efficient and flexible server provisioning is one of the most important criteria in measuring quality of cloud service. In order to save time and mitigate risk, images are set up as templates. And to provision a server, the cloud service provider replicates a selected image and then customize it to fulfill the server request. Therefore image set-up is vital to the efficiency and flexibility of cloud service. In this paper, we propose an optimization model of setting up images and mapping server request to correct image in server provisioning, with the purpose of minimizing total cost of setting up images, maintaining images and provisioning servers. After that, we compare the proposed approach with the vanilla image approach in server provisioning with numerical experiment.