Font Size: a A A

Research Of Video Stitching Technology Based On Mobile Environment

Posted on:2011-10-16Degree:MasterType:Thesis
Country:ChinaCandidate:L HanFull Text:PDF
GTID:2178360308455452Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
The stitching technology of video is a hot research. This technology has various application prospects in the areas of computer vision, image processing, such as the virtual reality, video conference, military surveillance, traffic guiding, and so on. Currently, besides ultra-wide-angle shoot we can also use the professional camera to create wide field-of-view photo. However, this type of system is expensive and complicated to be applied.Aim at the above problems about the key technology of video mosaic, the paper develops the research for video stitching technique. This research has important value and potential applications In order to reduce the number of camera and improve the speed of synthesis, this paper presents a technique to create panorama from video. This solution is based on mobile device, and this system has been optimized by the multi-core platform.The main contributions and features are listed as follows:1. Capturing the video sequence by mobile camera so the system does not need any special lens or hardware. The system is constructed by modules, and implemented with the Intel multi-core platform. As the moving mode of camera is adjustable, this system is convenient and easily expandable.2. Proposing a mosaicing algorithm based on key frames, which could reduce the redundancy for video sequences. This method adopts the correlation technique to select key frames, and uses parallel processing to optimize. So the selected key frames can be sufficient to describe all the video sequences of scenes.3. Implementing a feature extraction and matching modules based on improved SIFT algorithm. These modules are improved with parallel optimization. And then further filtrate the feature points with template operator, which enhances registration robustness and speed of synthesis. The experiment result demonstrates that the execution speed of improved feature extraction module is about 2.5 times faster than the original algorithm and this system can able to work under challenging conditions such as the translation, big rotation and so on.4. Exploiting the FPGA to implement the most computational-expensive feature extraction module. With the large amount of calculation of panoramic synthesis system, this system is divided into software modules and hardware modules. Then use Modelsim to achieve functional simulation, and verify the correctness of module operation.
Keywords/Search Tags:Video Panorama, Mobile Environment, Feature Extraction, Parallel Optimization
PDF Full Text Request
Related items