An Image Mosaics Method Based On Time And Frequency Domain Features

Image mosaic deals with two or a few overlapped images from the same scene, which mosaics them into a larger high qualified image. The automatic construction of large, high-resolution image mosaics is an active area of research in the fields of photogrammetry, computer vision, image processing, and computer graphics. This technology seamlessly mosaics a collection of partial overlapped images or video images into a large image with wider field of view. The emergence of this technology makes the image collection equipments more common, which can achieve satisfying image from images gotten by common video camera or digital camera. So researching and proposing an accurate and fast image mosaics method have extremely important practical significance.This paper introduces the conception of image mosaics, the classification of image mosaics methods and the present research situation. It summarizes the knowledge relative to image mosaics, which includes image getting methods, camera model, geometric transformation and so on. And it discusses all sorts of image mosaics methods, which include planar image mosaics methods, panorama mosaics methods, video image mosaics methods. Among them, it emphasizes on planar image mosaics methods, which include methods based on features, methods based on template and methods based on phase correlation.On this basis, it proposes an image mosaics method based on both time domain and phase domain, which aims at the shorts of present methods. It is a coarse-fine two steps method. Firstly, the coarse relationship between images is estimated by Fourier transformation in the frequency domain. The estimation of the angle becomes much easier as the introduction of the polar coordinate. Secondly, feature points are matched basing on this relationship. Finally, accurate transformation is derived with time domain features and images are stitched. Lab experiment proves that this method is efficient and robust, which stitches images seamlessly, without any manual intervention.