A Study Of Key Algorithms For Image Mosaic

With the rapid development of digital imaging technology, digital imaging equipment has been widely used. In some situations, images do not meet the need of application because of the limitation of them existing facilities, and then image mosaic technique was put forward. Image mosaic technique is an important research direction of image processing; it is widely used in digital video and computer vision, computer graphics, virtual reality, remote sensing image processing, etc. Based on the background of this, this paper used the digital image mosaic technology to generate the panorama image of large scale and wide visual angle.In general, image mosaic technology includes the following steps: image acquisition and image registration, image projection, image fusion, etc. If image was obtained with fish eye lens, the image must be corrected. Image registration is the difficulty and key problem in the field of image mosaic technology. It directly affects the success of image mosaic. Another key step in image mosaic is image fusion, it directly affect visual effects. Image projection used in displaying image, but realized in different manner. According to different requirement, the proper projection way may be chosen.By research each step of the image mosaic technology, the main work has been done as follow:1. Fish eye lens has the advantage of short focal length and large field, but fish eye image has severe distortion, so it must be corrected. Then contour extraction is the precondition of correction, contour extraction algorithm based on region growing is proposed in this paper. On the premise of image binary, image contour is obtained by means of an improved algorithm, and then image correction could be completed. It can overcome the condition that massive black pixels appear in the fisheye image circular outline area, and make the statistical algorithm fail.2. Image registration which is the basic component of image mosaic is transformed an unconstrained optimization problem. Parameter optimization model algorithm was proposed by means of quadratic form theory-- Orthogonal approximation algorithm. According to Properties of best-function, a direct method has been adopted without calculating derivative or gradient, only calculating function values. The advantages are eliminating of taking up large computer memory, only using a matrix to store information.Finally, I implement an image stitching system and demonstrate the steps with experimental pictures. The steps demonstrations of the image stitching system are realized by VC++ 6.0 and matlab 7.0. Simulation results show that proposed algorithms have advantage of practicality and effectiveness.