This paper presents a novel texture feature extraction method based on a Gabor filter and Three-Patch Local Binary Patterns (TPLBP) for Synthetic Aperture Rader (SAR) target recognition. First, SAR images are processed by a Gabor filter in different directions to enhance the significant features of the targets and their shadows. Then, the effective local texture features based on the Gabor filtered images are extracted by TPLBP. This not only overcomes the shortcoming of Local Binary Patterns (LBP), which cannot describe texture features for large scale neighborhoods, but also maintains the rotation invariant characteristic which alleviates the impact of the direction variations of SAR targets on recognition performance. Finally, we use an Extreme Learning Machine (ELM) classifier and extract the texture features. The experimental results of MSTAR database demonstrate the effectiveness of the proposed method.