The ColorDifferenceImage tells which regions in a image are closest to a given color. When a pixel in the Input image has a color close to C, the resulting pixel is light. Otherwise, it is dark.

ColorDifferenceImage(Input, C) begin   Output = NewImage(Width(Input), Height(Input), White)   for (x,y) in Input do     Set(Output, x, y, Distance(Color(Input, x, y), C))   return Output end

Once a color and a region are chosen, this image is calculated and is used to help determine the distribution of the given color that better approximates the region.

A test image and examples of its ColorDifferenceImage for several different colors.

StrokePositionsImage

The StrokePositionsImage algorithm takes the StrokeAreaImage and generates a distribution of points where the strokes will be placed. The desired characteristics in the distribution are the following:

• The concentration of strokes should be inversely proportional to the size of the strokes;
• It should not be possible to recognize any patterns in the distribution;
• It should be possible to control the concentration of points on the borders of the image;
• It should be possible to control the size of the largest region without a stroke.

The original implementation chose a Space-Filling curve dithering technique as a solution. We, however, developed a variation of the Floyd Steinberg Dithering algorithm, that spreads the error randomly in all directions. Furthermore, the error accumulated on each pixel follows the rule:

Therefore, the parameter p performs a gamma-correction on the image, controlling the concentration of strokes near the borders of the image. The parameter s is the largest length of pixels that can receive no stroke. The implementation is very simple and the results satisfactory:

	Original
	StrokeAreaImage
	s = 20 p = 3
	s = 30 p = 2
	s = 30 p = 3
	s = 30 p = 4
	s = 100 p = 5
Original Image, StrokeAreaImage and StrokePositionsImage for several values of p and s.