Geometric Filter

Transforms source image into a large number of geometric shapes, approximating the original image

Geometric Filter turns footage into a geometric approximation to simulate an artistic hand-drawn style. It does so by using Satori’s custom proximity matching algorithm, which gives it a bit of a human touch, as it can trace contours and similar regions.

The results are somewhat similar to what you get by using the so-called “style transfer” machine learning algorithms, but with much finer control over the result.

The plugin is aimed at artistic audiences who wish to simulate hand-drawn styles in music videos, visualizations, etc. and for VJs.

How it works

The filter works on a principle of finding similar pixels within a range and connecting them with geometric shapes. Following similarly colored pixels usually results in traced image contours, similar to drawing by hand. The process of picking pixels in the image can be tuned to be more predictable or more random.

Available parameters rundown

Shape parameters

The shape section allows you to tweak what kind of shapes are used for approximating the original image.

• Shape type – there are 6 geometric primitives available: lines, curves, triangles, squares, circles and rectangles.
• Count is the number of shapes used for drawing the image.
• When Recursion Steps are applied, the filter searches for continuous stretches of similar color in the same direction, which results in even more traced outlines.
• Size is the starting size of a shape (or the length of a line).
• Size Growth indicates how much the shape can grow in size when searching for similar colors. Effectively setting the size randomness.
• Line width affects just that.
• Fill toggles whether the shapes are filled or just their outlines are kept.
• Opacity sets the alpha of the shapes and gives the image a blurry quality.
• The final 3 parameters define the angle of the shapes. You can set the Starting Angle, Randomize it, and the Range specifies how far from the chosen angle the algorithm searches for a color match.

Image Matching parameters

Parameters in this section control the strategy of placing the shapes and color matching sensitivity.

• Accuracy describes how precise the filter should be in matching the original image’s colors. Decreasing it gives the output a more chaotic look and faster image processing.
• Fade Type decreases (or increases) the opacity of a shape with its size or its recursion step.
• Minimum and Maximum Luminosity limit what range of brightness from the image you wish to use. Useful for skipping too dark or too bright areas.
• Starting Position Strategy determines where the shapes are placed, which is especially important when animating. It gives the result more stable or wiggly look. The options are: Random, Stable Random or Grid.
• Grid Randomization ramdomly moves locations on the grid when using it as a starting position strategy.
• Random Seed (Evolution) – the randomness used for generating images always produces repeatable results. Use this parameter to change it is used for selecting can be used for animating image randomness.