Redundancy and Irrelevance Reduction

There are two basic concepts that enable us to compress images and videos: redundancy and irrelevance.

Redundancy is defined as wasted data. Meaning unnecessary additional bits spent to store information that would fit into less data bits.
For example imagine a tiny chest stashed in a large cardboard-box. It is a waste of space.
If we remove the redundant data from an image, it can still be displayed without any error or in other words: lossless. Completely removing redundancy is one of the main goals of image compression. This however is almost impossible as the effort needed to reduce redundancy grows as the coding rate approaches the entropy. (More about entropy in the chapter on entropy coding)

Irrelevance on the other hand is actual information in the picture that we purposely remove in order to lower the required bits for storing. But this removal comes at a cost: information lost cannot be recovered. The image coding process is now considered lossy. In order to justify the removal of information it is the job of the coder to select the kind of information which is least perceivable by the human visual system.

To sum up: irrelevance reduction is only used in lossy coding and redundancy reduction is the core concept of lossless coding but it of course used in lossy coding, too.