Time series are often defined as sets of sequentially-measured numerical values. With the development of sensors to measure various temporal indicators, time series have become more and more common. Such data is used to describe measured phenomena or to predict future values, using data mining methods. One crucial research topic when dealing with large time series is the reduction of dimension. Indeed, the often high dimensionality of the data may deteriorate the execution performance of sophisticated mining algorithms. Reducing time-series dimensionality is, therefore, a real stake of the data mining research for this data. In this paper, we propose to census the different methods from the literature that aim to reduce dimensionality on time-series. We both address the temporal-dimensionality reduction and, in the multivariate case, classic dimensionality reduction specifically applied to time series.