Similar pixels or image patches exist in different position inside an image or between different images, so that the fusion of similar patches can reconstruct the high resolution image. In order to mine the potential similar information further for the better reconstruction quality, the paper utilizes of the geometry transformation self-similarity of images, by extracting Zernike moments features within the image neighborhood to match and measure the nonlocal similar patches among the images. Then a super-resolution reconstruction algorithm for image sequences is proposed, through the collaborative weighting fusion of similar patches with geometric transformation. The experimental results prove the validity of the proposed method and its superiority in reconstruction effect.