Non-Local Means (NLM) algorithm is the best image denoising algorithms for its capability to better keep the image’s texture details and used widely in Remote Sensing image preprocessing area. However, the time complexity of the algorithm is very high due to its non-locality when searching similar pixels, which cannot satisfy the near real-time requirement in some specific applications. In order to resolve this issue, we designed and implemented a parallel NLM algorithm based on Intel Xeon Phi hardware, which has Intel's Many Integrate Core (MIC) architecture. In order to make full use of the computation resources, a dynamic load balance based collaborative parallel algorithm that can use both MIC and CPU resources is developed.