Study On Video Mosaic Construction Of Large Static Scenes

Automatic video mosaic construction is to form a large field of view including all information of a video sequence to solve the difficulty of full-view video observation caused by the limited view field of a camera. Automatic video mosaic construction has wide applications in civil life, remote sensing, survey of sea floor and underwater structures. For example, we need a real-time observation of the large field scenes in ocean floor survey, which is unable to be completed only by the general underwater cameras. Video mosaic technique is also of importance in construction of high-resolution images in remote sensing, medical micro-observation and virtual reality. Although panorama technology has been investigated extensively, the existing mosaic methods are still too time-consuming and inaccurate to meet the requirements of practical applications. Therefore, it is necessary to investigate and develop more effective mosaic techniques to satisfy the requirement of real-time video mosaic construction in some practical applications. Focusing on the above objective, we investigate the fast and accurate video mosaic algorithms for 1D scan pattern and 2D scan pattern, respectively. The following researches have been accomplished.(1) Study on video mosaic algorithm for 1D scan patternTo overcome the problem of registration error accumulation in traditional registration algorithms, this paper presents a fast and accurate video mosaic approach for 1D scan pattern. Different from the traditional video mosaic algorithm based on pair-wise local registration, the proposed new approach can directly compute the projective transform between the current input image and the panoramic image, thus effectively avoiding the accumulation and propagation of registration errors resulted from the concatenation of pair-wise transformations. The phase correlation based technique is employed to guide the extraction of feature points and matching process, thus reducing the time for corresponding point searching and improving the computation efficiency. The matching errors are eliminated using parameters of both phase correlation model and rigid motion model, which can avoid the effect of the local motions on the global motion estimation, so that fast and robust mosaic construction is implemented.(2) Study on video mosaic algorithm for 2D scan patternIn order to tackle the inconsistent issue of global video mosaics in registering video images from a 2D scan camera, we present a novel direct registration algorithm. When aligning the current image, it is not necessary to process a sequence of video frames, thus ensuring the real-time implementation. Moreover, the computation of the transformation between the current image and the panoramic image frame depends on both the temporal adjacent matching feature points and the spatial adjacent matching feature points, so the mosaic coherence can be achieved and the problem of registration error accumulation is overcome.The experimental results show that the proposed algorithm can construct seamless video mosaic quickly both for 1D scan mode and for 2D scan mode of a camera, and meet the requirement of real-time observation of large static scenes.