Polygon fragmentation method of distortion correction in computer image generating systems

• US 5,319,744 A
• Filed: 04/03/1991
• Issued: 06/07/1994
• Est. Priority Date: 04/03/1991
• Status: Expired due to Term

• Current Assignee: Intel Corporation
• Original Assignee: General Electric Company
• Inventors: Kelly, William A., Tackaberry, Michael W., Sims, Edward M., Quick, Lee T.
• Primary Examiner(s): Harkcom, Gary V.
• Assistant Examiner(s): JANKUS, ALMIS R
• Application Number:
  US07/679,807
• Time in Patent Office: 1,161 Days
• Field of Search: 395/119, 395/120, 395/125-130, 395/133, 395/136-139, 395/155, 395/161
• US Class Current: 345/647
• CPC Class Codes: G06T 3/00 Geometric image transformations in the plane of the image

Abstract

A method for distortion correction of computer-generated textured images maps vertices and texture coefficients from viewer space to projector space, so that environmental objects are pre-distorted upon the projection raster in order to appear in their proper form and perspective when the raster is projected onto a curved surface in viewer space, and viewed therefrom. Distortion correction is carried out by utilizing a piecewise-linear approximation for smoothly, continuously and closely approximating the required pre-curvature. The viewing space raster is subdivided into a number of triangles, within each of which a linear approximation is applied to the image mapping: the raster faces are first subdivided along the lines of a rectangular grid and the face fragments falling within each rectangle are then subdivided along the grid diagonal. This produces face fragments which are small enough so that a linear approximation provides an accurate transformation of each fragment. Distortion maps, typically computed off-line and stored in a database memory, are utilized to project the edges and texture modulation gradients from view space back into projector space, so that the edges and texture patterns can now be matched at intersections of adjacent triangles, in a manner to be substantially devoid of any abrupt changes, and thus be properly pre-distorted.

Claims

1. A method for pre-distorting a projection space PS video raster scene, formed of polygons having edges, to project upon a projection surface in a manner to be viewed as an undistorted VS video raster scene, having edge-defined polygons, by a viewer in viewer space VS, comprising the steps of:

(a) assigning a regular undistorted array grid with a multiplicity of row and column lines to divide the VS raster into an array of cells;

(b) arranging a diagonal line between one pair of opposite corners of each cell in the VS raster grid array;

(c) determining in the VS raster an ordered set of all points (VS crossing points) at which each polygon edge crosses any of the grid and diagonal lines;

(d) projecting each VS crossing point back into PS projection space to determine an associated PS space projected crossing point to which that VS crossing point will be transformed;

(e) creating a line connecting each pair of projected PS crossing points which correspond to a pair of VS crossing points connected by an interconnecting polygon edge in the view raster, with a plurality of PS connecting lines thus created enclosing an associated PS raster area; and

(f) matching textural area factors in each PS raster area thus formed to similar textural area factors in the corresponding VS area.

2. The method of claim 1, wherein step (a) further includes the step of forming the VS raster grid with rectangular cells.

3. The method of claim 2, wherein each cell is square.

4. The method of claim 1, wherein step (b) further includes the step of providing all of the VS raster grid diagonal lines in the same direction.

5. The method of claim 4, wherein all of the diagonal lines have a positive slope.

6. The method of claim 1, wherein step (d) includes the steps of:

storing a set of the equivalent VS and PS points for a predetermined VS-PS relationship; and

retrieving the stored set whenever that VS-PS relationship is used.

7. The method of claim 6, wherein step (d) further includes the step of storing pre-computed transformation point sets for a multiplicity of VS-PS relationships.

8. The method of claim 7, wherein the pre-computed, stored sets cover expected a selected range of VS-PS projection angle relationships.

9. The method of claim 1, wherein step (f) includes the step of matching the VS areal texture to a calculated PS texture at each of the set of crossing points.

10. The method of claim 9, wherein the view space areal texture is a function having both a numerator N_v and a denominator D_v and the projection space areal texture is a function having both a numerator N_p and a denominator D_p, and step (f) includes the steps of:

matching the PS numerator Np to the VS numerator Nv for each textured area; and

simultaneously matching the PS denominator Dp to the VS denominator Dv for that same textured area.

11. The method of claim 10, wherein the textured area matches are sufficiently close so as to prevent abrupt modulation changes at area boundaries.