Systems and methods for generating refined, high fidelity normal maps for 2D and 3D textures

摘要

Systems, methods and program storage devices are disclosed, which comprise instructions to cause one or more processing units to dynamically generate refined normal maps for 2D texture maps, e.g., supplied by a programmer or artist. Generally speaking, there are two pertinent properties to keep in balance when generating normal vectors comprising a normal map: “smoothness” and “bumpiness.” The smoothness of the normal vectors is influenced by how many neighboring pixels are involved in the “smoothening” calculation. Incorporating the influence of a greater number of neighboring pixels' values reduces the overall bumpiness of the normal map, as each pixel's value takes weight from those neighboring pixels. Thus, the techniques described herein iteratively: downsample height maps; generate normal maps; scale the normal maps to maintain bumpiness; and blend the generated scaled normal maps with generated normal maps from previous iterations—until the smoothness of the resultant normal map has reached desired levels.

主权项

1. A non-transitory program storage device, readable by a programmable control device and comprising instructions stored thereon to cause one or more processing units to:
(a) obtain a representation of a first two-dimensional image, wherein the representation comprises a first plurality of pixels, and wherein each pixel in the first plurality of pixels comprises a first plurality of pixel color values and a transparency value; (b) convert the first plurality of pixel color values into a luminance value for each pixel in the first plurality of pixels; (c) create a height map over the first two-dimensional image using the converted luminance values for each pixel in the first plurality of pixels, wherein each position in the height map corresponds to a pixel from the first plurality of pixels; (d) create a first normal map over the first two-dimensional image by calculating a normal vector for each pixel in the height map; (e) create a first blurred height map according to a first parameter; (f) create a second normal map over the first two-dimensional image by calculating a normal vector for each pixel in the first blurred height map; (g) scale the second normal map in accordance with a range of the first normal map; (h) blend the second normal map with the previously-created normal map to create a resultant normal map; and (i) repeat steps (e)-(h) iteratively, increasing the value of the first parameter with each iteration, until the resultant normal map has a desired level of smoothness.