Lossless algorithms - No information is lost. Repeated patterns in the image are utilized. RLE (run-length encoding) and PackBits algorithms are used.
- Other lossless algorithms are Huffman compression and LZW (Sempel, Ziv and Welch). LZW used with compressed tiff.
- Lossy algorithms - Based on assumption that high frequency data captured by scanner will not be perceived and can be discarded as noise.
- Most popular lossy algorithm is JPEG (developed by Joint Photographic Experts Group).
- Uses discrete cosine transform (DCT) a type of Fourier transform.
- JPEG can be very effective at times with insignificant loss of quality.
- However often can lead to excessive smoothing of detail, fringing or shade stepping in smooth areas.
- Postscript Level 2 can compress and decompress JPEG and LZW images.
- Recall that Photo CD images utilize subsampling (1 bit of color information kept for each 4 bits of lightness) which leads to significant file compression.
- Interpolation is used to restore as much lost color information as possible during decompression.