The Research Of View Manipulation For Image

View Manipulation for Image has always been the research focus in the field of Image Processing having attracting many researchers. View Manipulation for Image is to re-render the scene of the image under a new viewpoint by using Image Processing Technology. It is not only beneficial to make up for the lack of man’s imagination, but also can fix the image with a bad viewpoint. One current available method is to rebuild the accurate 3D coordinate of the scene by Image-based 3D Reconstruction technology and then re-project the scene to the new image plane corresponding to the new viewpoint. Another method is to drive the framework of the scene absolutely by hand.We present in this paper an efficient algorithm for manipulating the viewpoints of cuboid-structured images with moderate user interaction. Such images are very popular, and we first recover an approximate geometric model with the prior knowledge of the latent cuboid. While this approximated cuboid structure does not provide an accurate scene reconstruction, we demonstrate that it is sufficient to re-render the images realistically under new view points in a nearly geometrically accurate manner. The new image with high visual quality is generated by making the rest image region deform in accordance with the re-projected cuboid structure, via a triangular mesh deformation scheme. The energy function has been carefully designed to be a quadratic function so that it can be efficiently minimized via solving a sparse linear system. We verify the effectiveness of our technique through testing images with standard and non-standard cuboid structures.Without complex work of 3D reconstruction, the algorithm greatly reduces the complexity of the algorithm and can also work with only one input image. At the same time, compared with the method completely depends on user assistance, this algorithm can not only save a lot of manual operation during the processing, but also avoid errors caused by the subjective interactions. In mesh deformation process, we follow the conformal mapping principle and insist to keep the feature signed on the input image so as to draw the total energy equation that can be solved as a convex optimization with constraints, which can greatly enhance the speed of solving the equation.Finally, we can learn by experiment that the number of interaction on each input image is 7 on average and the energy equation of a 1024 x 768 image can be solved around 200ms. Compared with the result of Lightroom, our algorithm can avoid the obvious defects caused by Lightroom under the process of Image Warp. On the other hand, we can render the scene of image under a new viewpoint through view interpolation of some key viewpoint in real-time, which can facilitate the observation of the current scene under any sphere-viewpoint.