Image compression techniques can be grouped into lossless and lossy methods. Lossless methods preserve exact image data content. Compression ratios are of the order of 3:1. Lossy methods maintain an arbitrary level of image fidelity, and can produce compression ratios of 100:1 or more.
Compression symmetry is a measure of how much time and computational effort is required for encoding (compression) versus decoding (decompression). Quick image decoding (but slow encoding) is used when an image is transported to many different users, each of whom must decompress it. Quick image encoding (but slow decoding) is appropriate when many images are stored, but only rarely, if ever, viewed. This is the case in image archival applications (like storing the image of bank checks, or other paper records).
Image files can be divided into individual bitplanes or further processed using gray codes.
bitplane graycoded
plane size size
7 2,071 2,071 (same image)
6 4,160 3,462
5 5,993 4,952
4 7,880 7,214
3 8,487 7,461
2 9,166 8,040
1 9,709 9,172
0 9,791 9,701
total 57,257 52,073
original 47,948
The table shows the relative complexity of each plane, as measured by the compressed file size. However, there appears to be no advantage to encoded the bitplanes separately.
image files were encoded using gzip -9. Each image
file had a 128 byte header.
The fundamental form of predictive image coding is differential pulse code modulation (DPCM). The first pixel in a line is coded exactly as it is. For the next pixel, however, the value of the preceding pixel is subtracted from the current pixel, and the difference is encoded.
| 73 | 73 | 8 bits | |
| 48 | 48-73 | -25 | 6 bits |
| 76 | 76-48 | 28 | 6 bits |
| 56 | 56-76 | -20 | 6 bits |
| 83 | 83-56 | 27 | 6 bits |
| 40 bits | 32 bits |
In the above example, using a 6-bit code, there could be conditions where the difference between two adjacent pixels is greater than 31 or less than -32. This is called a code overload condition. Commonly a code (such as -32) is reserved to indicate the overload case. Then the following 8 bits contain the actual pixel value instead of the differential value.
Lempel-Ziv-Welch is the basis of zip and compress schemes.
usage: GZIP.EXE [-acdfhlLnNtvV19] [-S suffix] [file ...] -a --ascii ascii text; convert end-of-lines using local conventions -c --stdout write on standard output, keep original files unchanged -d --decompress decompress -f --force force overwrite of output file and compress links -h --help give this help -l --list list compressed file contents -L --license display software license -n --no-name do not save or restore the original name and time stamp -N --name save or restore the original name and time stamp -q --quiet suppress all warnings -S .suf --suffix .suf use suffix .suf on compressed files -t --test test compressed file integrity -v --verbose verbose mode -V --version display version number -1 --fast compress faster -9 --best compress better file... files to (de)compress. If none given, use standard input.
Maintained by John Loomis, last updated July 30, 1997