A new method for image mosaic is presented.Phase congruency algorithm is utilized to extract feature points. The proposed auto matching point optimizing method is used to realize the automatic selection of matching points. Levenberg-Marquardt optimization algorithm is used to estimate the transformation matrix between two images accurately. Ａnd, the seamless image mosaic is completed with smoothing algorithm. This method combines both feature points-based and optimization theory-based approach effectively, and it can make full use of image overlap information. This method can overcome the effect of noise and non-uniform illumination. Therefore it is more precise and robust than conventional algorithm. The experimental results illustrate that this method is very satisfied.