Image Drawn And Depth Of Extraction Technology Research

In recent years there has been increased interest, within the computer graphics and virtual reality communities, in image-based rendering (IBR) techniques. In contrast with the conventional geometry-based approaches of using 3-D models for creating new virtual views, IBR techniques use a collection of photographs as the underlying data representation, and render directly from these original images based on pixel interpolation or reprojection. These new techniques have many advantages over traditional geometry-based rendering, such as requiring only modest computational resources and achieving photorealistic virtual images with no additional operations. Camera calibration, depth maps recovering and constructing representation scheme are important issues in IBR, on which our research is concentrating.Camera calibration is the preliminary step towards deriving metric information from the images. We present a simple and practical self-calibration method of estimating the parameters of digital cameras for image-based rendering system. Depth maps recovering has long been one of the central research problems in computer vision. In the situation of image synthesis for IBR, we need not only precise but also dense depth maps. In order to meet the needs of this application, we propose several approaches to compute reliable dense depth maps, such as a feature-constrained stereo matching algorithm, a stereo matching algorithm based on gray segments and a stereo matching algorithm based on fuzzy identification, we demonstrate these techniques on synthetic and real image pairs, indicating that these approaches are feasible. The sampled representation scheme is the key problem in IBR, as the data structures decide how the reference images are collected and stored as well as how the virtual images are created. We develop a prismy panoramic model and discuss the acquisition and rendering of the datasets. We also describe an area-based method to create cylindrical panoramic images by stitching together a series of rotated overlapping photographs seamlessly.