Local Binary Pattern (LBP) is a kind of effective local texture description operator, which is widely used in many fields, like image processing, machine vision, etc, for its excellent rotation invariance. This paper mainly analyzes and improves the characteristics of LBP texture features, puts forward the Uniform Three-Patch LBP (UTPLBP) secondary statistic texture features and explores its application in the image segmentation. Regarding Brodatz texture image, UTPLBP operator is used to extract its texture features, and then the image is segmented by Support Vector Machine (SVM) algorithm. Experiment result demonstrates that UTPLBP operator can describe image texture information accurately and its secondary statistic texture features with strong robustness can distinguish different texture image. Compared with the traditional methods, the accuracy is better than ninety-three percent.