Computer graphics processing and selective visual display system – Computer graphics processing – Attributes
Reexamination Certificate
1999-09-10
2002-06-04
Brier, Jeffery (Department: 2672)
Computer graphics processing and selective visual display system
Computer graphics processing
Attributes
Reexamination Certificate
active
06400370
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention generally relates to computer graphics display systems, and more particularly, to techniques for applying texture to objects that are shown in the display. Even more specifically, this invention relates to a procedure for producing anisotropic texture using constant density object space stochastic sampling.
Many modern computer systems are able to display complex three-dimensional objects on display devices that are controlled by the computer systems, and commonly these complex objects are displayed interactively to allow the computer user to manipulate the objects. Well known graphics techniques for rendering these three dimensional objects such as Gouraud shading, hidden surface algorithms, clipping, filling polygons and coordinate transformations are used to generate the displayed object on a suitable device, such as a CRT video display that is controlled by the computer system.
Typically, in the operation of these computer systems, a polygon representation of the object is converted to a raster scanned image that is stored in a frame buffer. Usually, various parameter values, such as color, depth and translucency, are given for the vertices of the polygon, and in a process referred to as scan conversion, the computer uses the given values to compute values for these parameters for the pixels inside the polygon. The computed values are stored in the frame buffer at the resolution of the display device; and from that frame buffer, these values may be used to produce an image of the object on the display device.
Texture mapping is a commonly employed technique for adding detail in computer graphics rendering to achieve a high degree of realism in the rendered image. With this technique, image data needed to show the surface text for various objects, such as a road, the sky, a wheat field, or a brick wall, are stored in data arrays, referred to as texture maps, in the computer memory. When one of these objects is shown on the video display, data is obtained from the corresponding texture map and used to show the object on the display device with the desired surface texture.
In a common, simple texturing procedure, texture coordinates are specified for the vertices of each drawing polygon. These texture coordinates identify locations in the relevant texture map. As the polygon is rasterized pixel-by-pixel, these specified texture coordinates are used to determine, usually via an interpolation process, texture coordinates for each pixel. Then, as the image datum value for each pixel is being determined, the contents of the texture map are fetched, by means of the computed texture coordinates for that pixel, and used in the determination of that image datum.
Conventional texture mapping approaches used in real time and interactive graphic systems cannot achieve a quality texture mapping for shallow viewing angles of textured objects. This difficulty is demonstrated by a well-known example, referred to as the white line down the center of the road problem. In this example, the display system shows a white line down the center of a road. The white line is actually a series of white line segments, each segment is about ten feet long and five inches wide, and the line segments are separated by ten feet of black pavement. In showing this white line, the display system attempts to achieve two conflicting goals. The first goal is to keep the white line from disappearing as the viewer looks down the road, and the second goal is to prevent the white line from scintillating. Scintillation is manifested by the white line appearing for several segments (e.g., 50 feet), and disappearing for several segments (e.g., 50 feet).
Related problems are encountered when the display system attempts to show a building corner at a shallow angle. This building feature can be shown sharply when it directly faces the viewer, but the feature either scintillates or appears fuzzy when viewed at a shallow angle. As explained in greater detail below, these problems with conventional texturing systems are caused by the fact that the system does not account for the pixel's elongated footprint when the pixel is projected, or mapped, onto the texture map.
Texture maps are described in detail in U.S. Pat. No. 4,727,365, the disclosure of which is incorporated herein by reference. Generally, texture maps are organized into sets or series, with each set having a number of individual texture maps, referred to as Levels of Detail or LODs. Typically, each set has a base or highest resolution map, and successive maps in the set are reduced resolution versions of that base map. This base map is often referred to as LOD(
0
), and the successively lower resolution maps in the set are referred to as LOD(
1
), LOD(
2
), LOD(
3
), and so on. Commonly, each of these successive maps is a 2- to -1 reduction in each of the two dimensions of the previous map in the set. Thus, for example, if the base, or highest resolution, texture map in a set is 512 by 512 texels, then LOD(
1
) is a texture map that is 256 by 256 texels (individual elements of a texture map). Subsequent texture maps can continue to be formed until LOD(
9
) is produced, which would be a map that is 1 texel by 1 texel.
In the texturing process, the pixel is, in effect, projected onto a texture map, a process referred to as mapping the pixel into texture space. Then, two LOD values are determined based, respectively, on the width and the length of the projection of the pixel onto the texture map. The larger of these two values is taken as the effective, or actual, LOD value for the pixel.
After an LOD value is determined for a pixel, a texture value may be calculated for the pixel using any of a number of specific procedures. For instance, after the LOD value is calculated, the two texture maps with integer LODs that bracket the pixel's effective LOD value may be determined. Interpolation between adjacent texels in one or both of these LODs may be performed to calculate the value of the mapped pixel. Examples of current methods used to calculate pixel intensity include procedures referred to as nearest texel neighbor, bi-linear interpolation using texels in one of the integer LODs, and tri-linear interpolation uses bi-linear interpolation on two adjacent LODs and linearly blends the results from each of the bi-linear interpolations.
These forms of texture mapping are said to be isotropic, since the LOD value for the projected pixel is assumed to be equal along both the length and the width of the projection of the pixel onto the texture map. This isotropic assumption produces excellent results for pixels that project substantially orthogonally onto a textured surface. However, this assumption may produced some errors for pixels with elongated projections onto the texture map, for instance as shown in FIG.
1
.
Anisotropic texture approaches that account for a projected pixel's elongated shape in texture space have been implemented, but only for non real time systems. Such approaches fall into two broad categories: 1) methods for convoluting the pixel's projection in texture space with the texture values, and 2) methods using the storage of pre-processed and pre-filtered textured maps. The former methods can be very accurate but computationally very expensive. The latter methods require less computations than the former, but are less flexible, less precise, and are memory intensive.
Because of an increasing demand for more and more different textures maps to be available in a graphics display system for real time and interactive applications, the texturing approaches that are less memory intensive are more desirable.
U.S. Pat. No. 5,651,104 for “Computer Graphics System And Process For Adaptive Supersampling” addresses the problem of anisotropic texture mapping.
FIG. 1
shows the projection of a pixel onto a texture map. This patent suggests sampling regularly along the substantially longitudinal axis of the pixel's projection onto the texture map. Several difficulties arise, though, when samp
Economy Richard
Krupnik Stuart
Lee Harry
Brier Jeffery
Cunningham G. F.
Intel Corporation
Scully Scott Murphy & Presser
LandOfFree
Stochastic sampling with constant density in object space... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Stochastic sampling with constant density in object space..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Stochastic sampling with constant density in object space... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2898957