In this correspondence, we deal with the problem of channel estimation in amplify-and-forward (AF) wideband cooperative relay-based networks. Two types of frequency domain channel estimation techniques are proposed and analyzed. First, a training based technique is presented for which an optimal pilot placement and power allocation strategy is described. Second, hybrid techniques are introduced in which both training as well as channel output correlation information is utilized for channel estimation. A theoretical performance study of the proposed algorithms is presented and closed-form expressions for the mean squared channel estimation error are provided. The presented theoretical analysis is verified via extensive Monte Carlo simulations.