ImageFlow: Real-Time Image-Based Rendering

深度图像被用作本项目正在构建的高性能图形引擎中的渲染原语。深度图像是一种二维图像，除了颜色和其他属性外，还包括相对于视点的深度。从少量的深度图像中，每个图像都代表了一个单一视点的任意复杂场景，可以计算出原始视点附近的任何视点的图像。这种基于图像的渲染方法在真实感方面具有优势，并且在生成深度图像和将其重新投影到新视点的处理器之间分配渲染计算。该项目正在使用现有的PixelFlow图形引擎来测试基于图像的渲染想法。在项目的第一部分中，将研究获取和预处理深度图像的算法。与此同时，PixelFlow机器将在图像采集和存储方面得到增强。这将允许基于图像的渲染软件在PixelFlow机器上实时运行，执行以下操作：确定用作深度像素源的参考图像，将参考图像像素扭曲到屏幕空间，评估每个像素的颜色，面积和其他参数，以及混合像素以形成最终的参考图像。在项目的后期阶段，可能会为其中一些操作开发和评估硬件支持。

Depth images are used as rendering primitives in the high-performance graphics engine that is being constructed in this project. A depth image is a two- dimensional image that includes depth relative to a viewpoint, in addition to color and other properties. From a small number of depth images, each of which represents an arbitrarily complex scene from a single viewpoint, an image can be computed for any viewpoint within the neighborhood of the original viewpoints. This image-based rendering approach offers advantages in realism and distributing the rendering computation among processors that generate depth images and those that reproject them to new viewpoints. The project is using the existing PixelFlow graphics engine to test the image- based rendering ideas. During the first part of the project, algorithms for acquiring and pre-processing depth images will be investigated. At the same time, the PixelFlow machine will be enhanced for image acquisition and storage. This will permit software for image-based rendering to run in real-time on the PixelFlow machine, performing the following operations: determination of reference images to be used as sources of depth pixels, warping of reference image pixels to the screen space, evaluation color, area, and other parameters for each pixel, and blending of pixels to form the final reference image. In later stages of the project, hardware support may be developed and evaluated for some of these operations.