In cloud computing, a fundamental management problem with the Infrastructure as a Service model lies in the efficient embedding of computing and networking resources onto distributed virtualized infrastructures. This issue, usually referred to as the Virtual Network Embedding (VNE) problem, has been well studied for a single Cloud Provider (CP). However, wide-area services delivery may require to embed heterogeneous resources over multiple CPs. This adds more complexity and scalability issues, since the Virtual Network Requests (VNRs) embedding process requires two phases of operation: the multicloud VNRs splitting, followed by the intracloud VNR segments mapping. This paper addresses the problem of VNE across multiple CPs by proposing a VNRs splitting strategy which aims at improving the performance and QoS of resulting VNR segments. An Integer Linear Program (ILP) is used to formalize the splitting phase as a maximization problem with constraints. Subsequently, in order to minimize the overall delay, a multi-objective intracloud resource mapping approach formalized as a Mixed-Integer Linear Program (MILP) is adopted. Simulations with the exact method show the efficiency of the proposed strategy based on several performance criteria. In particular, the acceptance rate and the delay are respectively improved by 15.1 and 18.5 percent, while preventing QoS violations.