The Research of Intermediate View Synthesis Algorithm Based on NSCT and Image Fusion
|School||South China University of Technology|
|Course||Signal and Information Processing|
|Keywords||Intermediate view synthesis NSCT Image Fusion Deblocking|
With the development of information and communication technology, the traditional 2D video planar image description has been unable to meet the demand for interactive and immersive 3D video technology can engender a strong sense of immersion and spot impulses, becoming multimedia important research content of the communication. Users can choose angles and viewpoints freely through the 3D TV and 3D visual experience. Due to uneven sampling shooting and the limit of the amount of storage hardware, the images obtained by physical means often can not meet the needs of the user, so the view synthesis technology came into being. View synthesis (View Synthesis) also referred to as the image synthesizing or image deformation means through two or more camera viewpoint on the same image of the scene, directly calculate the image of the scene on each of the other camera viewpoint or gaze direction. The synthesis of the middle view is a view synthesis techniques, it is mainly used in the 3D TV, a virtual video conferencing. Intermediate view synthesis technology rendering technique based on three-dimensional geometric model (Modeling-Based Rendering, MBR) and image-based rendering technology (Image-Based Rendering, IBR) two, MBR technology the need for the physical scene modeling rendering speed depends on the complexity of the scene, in real-time is not high; IBR by the real image, to extract useful information to the synthesized image does not require modeling of the scene, the drawing speed is higher. The IBR based method, proposed an intermediate view synthesis method. The method first registration image, and then transformed into NSCT domain in the transform domain of the image fusion process, and the fused image to a filtering process, removing the effect of the sub-block, and ultimately obtain the desired composite view. This paper studies content: 1) study the feature point-based image registration techniques; 2) research-based the NSCT transform domain image fusion technology; 3) study the post-processing techniques to remove the effects of the sub-block; 4) of this article The algorithm for the simulation, and the simulation results are analyzed. The advantage of the algorithm: 1) NSCT transform to solve the wavelet transform sub-sampled bring the spectrum aliasing problem; 2) transform domain image fusion method, filling part of the void, and achieved good results. The algorithm has the following disadvantages: 1) the synthesis of non-parallel view research; 2) the lack of scientific the synthetic image quality evaluation system; 3) the efficiency of the algorithm needs to be improved.