Non-photorealistic image rendering with a labyrinthine tiling

Amelia Carolina Sparavigna

Photorealistic processing of images is often required for scientific and forensic imagery, where its  processing algorithms can remove noise or enhance object outlines in the image scene. However, it is not obvious that a photorealistic processing is always necessary or useful. Let us consider for instance the hand-drawing, which is clearly a non-photographic imagery. The idea that these illustrations  can be better to explain a scene than photographic plates is quite natural. Hand-drawing helps to illustrate complex phenomena, omitting details and proposing only fundamental objects, sometimes with a symbolic representation.

The goal of the non-photorealistic rendering (NPR) (for a recent literature survey,1-4) is the development of algorithms for generating or processing images that embody the following qualities: emphasis of selected features, suppression of unimportant details, use of stylization to suggest emotional structures. Some NPR algorithms are devoted to produce pseudo hand-drawing images (5). Other techniques are giving interesting aesthetic results (6) working with algorithms for visualizing the vector fields. The vector field visualization, used also for scientific purposes (7,8), is mainly based on line integral convolutions (LICs). LIC and related techniques use stylization to suggest the structure of the vector field, as an hand-drawing stylization of the field can do.

Here we can see the results of an image processing based on a statistical approach, resulting in non-photorealistic rendering, different from vector field visualization, or from other techniques such as mosaic or tessellation textures. It is based on the use of statistical parameters and gives on the processed image a labyrinth tiled texture, which is not governed by casualty but able to follow the outlines of the objects in the image scene.

The rendering  is a random pixel tone generation performed with two competing statistical requirements. This is giving a peculiar pattern on the image, producing a tiling of the image with a labyrinth-like texture. As in many physical and chemistry phenomena, where competing conditions are acting on the system, complex patterns are displayed (9-11). It is quite usual to encounter stripes, fingerprints or labyrinths in the optical investigations of solid surfaces or of thin liquid crystal films. Competing conditions give two or more possible local orientations or configurations of the material, revealed by complex patterns in microscopy observations.

The algorithm is based on a statistics manipulations of the image. The statistical approach means that decisions and choices of the algorithm are based on statistical parameters such as mean values, variances and higher moments (12-14). For a detailed discussion of the algorithm, see  http://arxiv.org/abs/cs/0609084,

The stylization here proposed is able to give in certain circumstances, a better appreciation of a scene. An example of the algorithm applied to the picture of a tree (see below), can be useful to stress this possibility. In the rendered image, the details of the small branches of the tree are suppressed but the shape of the main branches is strongly enhanced by the labyrinth tiling. In this case then, the morphology of the tree is put in evidence by the rendering: it seems that the growth of the branches in the image is following the labyrinthine pattern. The intrinsic geometric structure (the tree) in the image is favoured by the stylization of the rendering algorithm.

As we have argued at the beginning, one of the goals in non realistic rendering is to stimulate sensations, with a certain enhancing of patterns in the image scene. Whether the rendering algorithm turns out to be good or not for this purpose is of course a rather subjective conclusion.

REFERENCES

[1] F. Durand, An Invitation to Discuss Computer Depiction, Proc. Int. Symp. On Non-photorealistic

Animation and Rendering, NPAR02, ACM Press, pp. 111-124 (2002).

[2] M. Tory and T Moller, IEEE Trans. Visualization and Computer Graphics 10, 72-84 (2004).

[3] T. Strothotte and S. Schlechtweg, Non-Photorealistic Computer Graphics: Modeling, Rendering, and

Animation, Morgan Kaufman Publisher, San Francisco (2002).

[4] C. Reynolds, Stylized Depiction in Computer Graphics, an Internet report (2006)

http://www.red3d.com/cwr/npr/, 2006.

[5] I. Buck, A. Finkelstein, C. Jacobs, A. Klein, D. H. Salesin, J. Seims, R. Szeliski, and K. Toyama,

Performance-driven hand-drawn animation, Proc. Int. Symposium on Non-photorealistic animation and

rendering, NPAR00, ACM Press, pp.101-108 (2000).

[6] C-M. Wang and J-S. Lee, J. Information Science and Engineering 20, 923-948 (2004).

[7] A.Sparavigna, A.Sanna, B.Montrucchio and A.Strigazzi, Liq. Cryst. 26, 1467-1478 (1999).

[8] B. Montrucchio, P. Montuschi, A. Sanna, and A. Sparavigna, Computers & Graphics 25, 847-855 (2001).

[9] J.P. Gollub and J.S. Langer, Rev. Mod. Phys. 71, s396-s403 (1999).

[10] C. Bowman and A. C. Newell, Rev. Mod. Phys. 70, 289-301 (1998).

[11] A. J. Koch and H. Meinhardt, Rev. Mod. Phys. 66, 1481-1507 (1994).

[12] B.Montrucchio, A.Sparavigna and A.Strigazzi, Liq. Cryst. 24, 841-852 (1998).

[13] B.Montrucchio, A.Sparavigna, S.I.Torgova and A.Strigazzi, Liq. Cryst. 25, 613-620 (1998).

[14] A.Sanna, B.Montrucchio and A.Sparavigna, Pattern Rec. Lett. 20, 183-190 (1999).

Here in the following a gallery of images obtained after tiling.

Copyright © 2009 Amelia Carolina Sparavigna