Abstract

Data compression involves generating a compact description of information from a representation that may contain redundancies either of exposition or of interpretation. The process of generating a compact description involves understanding how information is organized in data. One way this can be done is by understanding the statistical characteristics of the data. If the data sequence is independent, we could treat the data as a sequence of independent, identically distributed (iid) random variables with a particular probability distribution. If the data are correlated, we can develop models for how redundancies could have been introduced into a nonredundant information stream. The study of data compression involves ways to characterize the structure present in the data and then use the characterization to develop algorithms for their compact representation. We will look at a number of different ways in which commonly used sources of data, such as text, speech, images, audio, and video, can be modeled and the compression algorithms that result from these models.