When you decrease an image’s color depth, you must select a color reduction method. Depending upon the color depth you select, two or three of these methods are available:
Nearest Color Replaces the original color of a pixel with the color in the newly generated palette that is closest to the original RGB value. This method eliminates dithering and produces a high-contrast image. Simple graphics may look better with this method.
Error Diffusion Replaces the original color of a pixel with the most similar color in the palette, but spreads the discrepancy between the original and new color to the surrounding pixels. As it replaces a color (working from the top left to the bottom right of the image), it adds the “error,” or discrepancy to the next pixel, before selecting the most similar color. This method produces a naturally looking image and often works best for photographs or complex graphics.
Ordered Dither Adjusts adjacent pixels so that two colors give the illusion of a third color and intermingles pixels to produce patterns based on a known palette. Use this method to give the image the appearance of containing more colors than it actually does. Images may appear to be composed of cross-hatches and dots and may have distinct patterns of light and dark areas.
When you decrease an image’s color depth, you must select a palette option. Depending upon the color depth you select, two or three of these methods are available:
Optimized Median Cut Measures and ranks each color by how frequently it occurs in the image, then uses the most frequently occurring colors to determine the new palette. Even if your image contains fewer colors than the palette that is generated, this method may not represent each color exactly. The method, therefore, is not as accurate as the Optimized Octree method, but it is better at weighting color importance.
Optimized Octree Creates an eight-level “tree” with eight branching nodes totaling 256 individual “leaves” or colors to determine the new palette. This method attempts to reproduce each color in the original image, so if your image contains fewer colors than the palette that is generated, every color in the image is represented. This method is faster and more accurate than Optimized Median Cut, but it is not as good at weighting color importance.
Windows Changes each pixel’s color to the nearest color in the Windows palette.
Standard/Web-Safe Palette Changes each pixel’s color to the nearest color in the standard Web-safe palette. Use this method to create images for the Web that can be viewed without color distortion on most monitors.
Understanding Color and Color Models
How Monitor and Print Colors Differ
Making a Palette Color Transparent