A Survey Of Image Mosaic Techniques

Nowadays, images with wide field of view and panoramic images are used in more and more applications. They can be produced by some special facilities (such as push-broom camera or fish-eye camera), whose disadvantages are higher price and more complex manipulation. For this reason, how to get such images with normal cameras has attracted many attentions. This problem can be resolved by image mosaic technique, a branch of computer vision and image processing, which mosaics a collection of partial overlapped images into a large image with wider field of view.This paper proposes a similar-curve-based algorithm to get cylindrical panoramic image , which is wildly used in the applications of image-based virtual reality . In the proposed algorithm, Zernike moments are presented to solve for the translational registration, rotational registration .First, some feature points are extracted based on B-Spline wavelet transform from two images. Noisy edge information is vanished across multiscale edge detection. The last, a recursive scheme based Zernike is improved for deriving rotation angle of two overlapping images, The proposed method solve the following problems: (1) require a great deal of overlap between images or (2) or if they work for restricted sub-problem (translation, rotation or zooming) they would not work for the others. The proposed method has been tested and worked well on many images of static scene taken with a hand-held digital camera.