In contrast to the device-dependent colour spaces RGB or CMYK, the CIELab colour space describes colour objectively, i.e. the colour values are thus clearly and unmistakably defined. The Lab color space separates the brightness (L) from the two color coordinates a- (red-green) and b-axis (yellow-blue). In addition, this color space includes all color impressions perceptible by the human eye, which are represented in it with approximately equal visual distance. An equal distance between two points results in a nearly equal sensation for the color difference of these two colors everywhere in this color space.

Figure 1: Schematic display of the Lab Color Space

Light inks have a lighter concentration. It is usually only a dilution of the respective base colour. They are mainly used to achieve pastel shades on large areas in order to partially solve the problem of visible graininess caused by dithering.

The process of measuring and correcting for a device's inability to see or reproduce a straight line of tones from black to white.

The process of Linearization is most commonly used to ensure an image setter reproduces the same halftone dot values predicted by the imaging software.

The work of Abraham Lempel and Lempel Ziv formed the basis for a large number of coding methods, which are generally referred to as LZ methods. Based on the original methods LZ77 and LZ78, a wide variety of derivatives have been developed over the course of time. The method used in PDF is LZW.

Functioning of LZW: LZW is based on LZ78 and was modified by Terry A. Welch in 1984. LZW is probably the most widely used lossless algorithm, which is used in TIFF, GIF, PDF and also PostScript (from level 2) to save data more compactly.

Like RLE, LZW can also be applied to any form of data, regardless of whether it is text or image data. Compression and decompression take place at the same speed (= symmetrical compression), as the dictionary is created at runtime during both encoding and decoding.

LZW is based on a dictionary. First of all, the algorithm constructs a data dictionary with a maximum of 256 entries of all uncompressed information. Then, in contrast to RLE, the data stream is not searched for "runs" of identical information values, but for patterns (substrings) in the data stream. From the pattern, encodings are calculated that represent the compressed data. If an already known pattern occurs again in the image data, the existing coding is used and the pattern is written into the compressed data. The word "will" can be made up of the known part "w" and the as yet unknown part "ill", resulting in a new pattern. The data dictionary can contain up to 2¹² entries.

A special feature of LZW is that the lexicon does not have to be saved. This further improves the compression rate. Elements that are stored in the lexicon - for example: the first 128 characters of the ASCII code - are stored as 1-byte values and are therefore available for compression and can be restored from them.

Difference calculation: In order to compress data better, the method of difference calculation (= differencing) is often used. With this type of calculation, it is not the value (information content) of the next pixel that is saved, but only the difference to the original pixel. In halftone images in particular, directly adjacent pixels differ only slightly, which means that differential calculation results in data reduction, which can of course also be restored completely correctly. For 1 to 8-bit data, the differences between the pixel values are calculated directly; for RGB data, only the differences in the individual channels R, G, B are calculated and used for compression. This limits the compression rate somewhat.

The Predictor Key can be used to tell the decoder whether normal LZW compression (Key=0) or extended differential compression should take place. The predictor key for TIFF is always 2. Predictor keys above 10 are used for the successor format to GIF (PNG).

Summary: LZW is just as efficient as RLE for 1-bit data. The compression rates for 8- to 24-bit images are usually more efficient with LZW, especially if noise suppression has taken place in the images. However, more effective lossless data compression can be achieved with "flat" color images such as cartoons or technical plans. With this type of image in particular, sharp edges and pure surfaces must be retained.

Technically speaking, LZW is one of the lossless, symmetrical compression methods, whereby compression is performed logically. In PDF, LZW is addressed via the LZWDecode-Filter. It is generally possible to parameterize the filter, but this is not usually provided for in the user interface of graphic, layout and PDF editors.