Color reduction

Color quantization or color reduction are called quantization processes in computer graphics , which reduce the number of colors in a raster graphic .

Color reduction is used in order to benefit from the smaller storage space of a graphic file with a lower color depth . Indexed colors in particular enable relatively compact files. Color reduction is becoming less important as today's graphics cards and screens support true colors .

Procedure

There are different algorithms to determine the best colors (representatives) for the image to be processed. These include: ${\displaystyle n}$

• Uniform quantization
• Popularity algorithm
• Octree
• Median cut
• NeuQuant ( self-organizing card )
• Generalized Lloyd algorithm (Linde-Buzo-Gray)

Improvement of the picture impression

The reduction of the colors creates edges that are perceived preferentially by the eye ( Mach stripes ) and disturb the image impression. So-called dithering smears the transitions through selective noise and subjectively improves the image quality.

Examples

• 

  Original image (true colors)

• 

  Reduced to 16 colors, without dithering

• 

  Reduced to 16 colors, with dithering

The color reduction is necessary when saving in a graphic format that only supports a lower color depth than the original image. For example, the GIF format supports 256 different colors. Grayscale images with a color depth of 8 bits can be saved without loss, but color photos with a color depth of 24 bits per pixel must be reduced to 256 colors before being saved.

literature

• James D. Foley et al. A. Computer Graphics: Principles and Practice , 2nd Edition, Addison-Wesley 1995, ISBN 02-018-4840-6