P Example-based Approaches Image-based Modeling, Rendering, and Lighting Example-based Super-resolution

objects offer resolution independence over a wide range of scales. With this approach, object boundaries remain sharp when we zoom in on an object until very close range, where faceting appears due to finite polygon size (see Figure 1). However, constructing polygon models for complex, real-world objects can be difficult. Imagebased rendering (IBR), a complementary approach for representing and rendering objects, uses cameras to obtain rich models directly from real-world data. Unfortunately, these representations no longer have resolution independence. When we enlarge a bitmapped image, we get a blurry result. Figure 2 shows the problem for an IBR version of a teapot image, rich with real-world detail. Standard pixel interpolation methods, such as pixel replication (Figures 2b and 2c) and cubic-spline interpolation (Figures 2d and 2e), introduce artifacts or blur edges. For images enlarged three octaves (factors of two) such as these, sharpening the interpolated result has little useful effect (Figures 2f and 2g). We call methods for achieving high-resolution enlargements of pixel-based images super-resolution algorithms. Many applications in graphics or image processing could benefit from such resolution independence, including IBR, texture mapping, enlarging consumer photographs, and converting NTSC video content to high-definition television. We built on another training-based super-resolution algorithm and developed a faster and simpler algorithm for one-pass super-resolution. (The one-pass, example-based algorithm gives the enlargements in Figures 2h and 2i.) Our algorithm requires only a nearest-neighbor search in the training set for a vector derived from each patch of local image data. This one-pass super-resolution algorithm is a step toward achieving resolution independence in image-based representations. We don’t expect perfect resolution independence—even the polygon representation doesn’t have that—but increasing the resolution independence of pixel-based representations is an important task for IBR.