Image mosaic technique is used to create vast scene images and panoramic images. It has widespread use value in the cultural heritage protection and virtual reality field. As we know, there are many methods of image mosaic, such as feature points and binary image region matching, etc. However, the traditional processing methods always ignore speed or precision or practical scale; that is to say, they are unlikely to give attention to two or more things. In order to achieve much better and optimal effect of image mosaic, we propose a model of least square adjustment, which is similar to the strip block adjustment in traditional Aerophotogrammetry. This model determines the relatively spatial relationship of the image facade texture. Experimental results show that it is a robust, efficient and practical image mosaics algorithm. Furthermore, this algorithm expands the applicable scope of the transitional the strip block adjustment in Aerophotogrammetry.