Cloud computing centers enable to provision diverse virtual machine (VM) in an elastic and scalable manner. VMs are created by instantiating VM images that encapsulate the required applications or services with their support environment. The multi-gigabyte VM images are hosted in the central image repository, while hosting centers are geographically dispersed. Thus, to fetch a GigaByte sized VM image from a repository over the Internet is time consuming. To address this problem, we first take advantage of high degree of commonality of VM images, and divide the VM images into fine-grain chunks by designing a mixed chunk scheme. And then, based on chunk-level sharing among different images, we propose a novel VM images acquisition scheme based on multi-point collaboration (MPCA), which minimizes latency in fetching VM images. Our experiments show that our proposed scheme can largely reduce request service time. Moreover, it can efficiently decentralize network traffic load, reduce the contention result from additional traffic introduced in the network, especially when traffic load is heavy.