The distinguishing geometric features determine the success of point cloud registration. However, most point clouds are partially overlapping, corrupted by noise, and comprised of indistinguishable surfaces, which makes it a challenge to extract discriminative features. Here, we propose the Neighborhood-aware Geometric Encoding Network (NgeNet) for accurate point cloud registration. NgeNet utilizes a geometric guided encoding module to take geometric characteristics into consideration, a multi-scale architecture to focus on the semantically rich regions in different scales, and a consistent voting strategy to select features with proper neighborhood size and reject the specious features. The awareness of adaptive neighborhood points is obtained through the multi-scale architecture accompanied by voting. Specifically, the proposed techniques in NgeNet are model-agnostic, which could be easily migrated to other networks. Comprehensive experiments on indoor, outdoor and object-centric synthetic datasets demonstrate that NgeNet surpasses all of the published state-of-the-art methods. The code will be available at https://github.com/zhulf0804/NgeNet.