Stroke-based NPR

Applications for stroke-based NPR. The first column is using very few number of streamlines to draw the image(around 2200). And the second column is using a lot of streamlines (over 40000) to draw the image.

Problems I can see from the applications:

1. Quality of ETF. Might be made up by post drawing. E.g. losing sharp corners.

2. Shape, thickness, and the length of each stroke are not easy to choose. When is good for visualization?

3. The order of the streamlines? Based on Hertzman? From the longest to the shortest?(to be tested)

4. The color choice of each stroke? Constant colors and variant colors and textured strokes.

5. How to enhance stroke-based NPR based on current choices such as:color, directional, structure?(Some choices to be proposed and should be balanced between ...).

6. Previous ETF papers mentioned the differences between their methods and LIC. I am not sure if it's a drawback for ours.

7. How far the local controls can go? What's the goal of our styles?

Advantages:

1. A way to abstract an image

2. Directional and spatial advantage

3. Local adjustment

4. Speed??

5. Automatic

6. A new way for placement of strokes