Encyclopedia of Graphics File Formats

Compression Demo

There are many methods of compressing graphics files. This page attempts to demonstrate how lossless and lossy compression compare. The test data for this demonstration is a 24-bit color image. The bar graph in the last column of these tables is intended to provide a visual comparison of the relative file sizes.

First, we compare several lossless techniques:

Compression Method	Resulting Size	Resulting Percent	Visual Comparison
Uncompressed image	4,261,256 bytes	n/a (100%)
24 bit PNG	2,081,345 bytes	52%
8 bit PNG	620,254 bytes	86%
JPEG 100%	967,111 bytes	78%

Note: Comparing 8 bit PNG to the 24-bit raw data is misleading. Although PNG is a lossless compression method, using only 8 bits of color represents a loss of 16 bits per pixel. Note also: strictly speaking, the JPEG method, even at 100%, is a lossy compression method.

Lossy compression can have a much more significant payoff, but always at the expense of some image quality:

Image	Compression Method	Resulting Size	Resulting Percent	# Colors	Visual Comparison
	JPEG 100%	967,111	n/a (100%)	287,723
	JPEG 75%	154,091	84%	302,429
	JPEG 50%	99,819	90%	282,068
	JPEG 25%	65,376	93.2%	207,908
	JPEG 1%	24,173	97.7%	90

Note: the uncompressed 24-bit image uses 130,851 colors.

Selecting the best compression method for any particular image depends on many factors: the number and distribution of colors, the image quality required by your application, and the nature of the tools that must manipulate or display the image. Running comparisons similar to the ones shown above on your real images will provide a better indication of the payoff for your particular application than we can provide in the general case.

For further information on compression of all kinds, see the comp.compression newsgroup FAQ.

Copyright © 1996 O'Reilly & Associates, Inc. All Rights Reserved.