Results after 10000 iterations and smoothing

The skeletons are obtained based on region boundaries after segmentation.

Problems for tracing the streamline based on ETF

Use a sphere to pop up each pixel on each streamline or just pop it up.

Problems: The length for each streamline is too short. How to get a long streamline?

Perturb the normal based on chosen texture and results based on NPAR2007 paper

After finding ETF and do DoG.

Texture

Problems for perturbation: it is not seamless at the boundaries.

Some intermediate results

Original

Lighted results(2000 iterations)

Connected skeletons

What am I doing now?

1. Segment the image.
2. For each region, get a better connected skeleton.
3. Normalize.(0 ~1 from boundary to skeleton)
4. Choose a function for popuping.(y = sqrt(1-x * x) * scale or else)
5. Do the relaxation and also keep the values from 4.
6. Do the light calculation.
7. Perturb the normal based on chosen textures(based on bump surface).
8. Find the ETF and trace the streamline for high frequency areas.
9. For each found streamline, popup it up as a line or popup it up as a stroke.

No good results come out yet. Later soon....

Hairy

smoothed

unsmoothed

Movie for water

Links for download:

More varieties for water surface

Fixed viewing and the changing light The center has higher probabilities for placing the sparks. The farer, the more the number of sparks. And the farer, the smaller the size of sparks.

Sparks without mixture with original colours

Control the size of the sparks based on the distance between the eye and the water vertex

sparsely scattered

densely scattered

Sparky water surface

Segmentation







Modified
Original