In this paper, we propose a new algorithm for optimizing surface texture mosaics when texturing 3D model reconstructed from multi-views of objects from real world. We start with the 3D geometry acquired by structure from motion (SFM) approach. Then, the photographic images for acquiring 3D structure are used as texture data and mapped precisely on the surface. Inspired by law of Matthew Effects, we implement an iteration algorithm for automatically sampling a few images as texture maps instead of using the whole set of input images. Imposed on visibility and texture area constraints, texture mosaics are efficiently regrouped in the iteration framework. A great deal of seams between different texture mosaics where texture maps belong to different views are relieved after performing our optimization method. The implemented texturing system and its sufficient experimental results demonstrate the practicability and efficiency of the proposed approach.