This dissertation aims to analyze and compare trend filters, applying them to trend-following strategies. The proposed methodology can help in decision making for the construction of investment strategies. Considering the search in the literature for techniques of extracting trends that avoid overfitting, this work will analyze different filters: L1 filter (Kim et al., 2009), moving average filters, Hodrick-Precott filter (Hodrick et al., 1997) and the Kalman filter (Kalman, 1960). For a database consisting of stock exchange ETFs (Exchange Traded Funds) of emerging market stock indices, the presented methodology proposes to comparatively evaluate the performance of trend-following strategies when applying each of the filters. The filters are comparable, since they will be applied to the same strategies, the same assets and with the same computational resources. Considering recent analyzes and good performance, emphasis will be placed on the L1 filter, which is a nonlinear filter, different from the others used in this work. The results of this dissertation indicate that the L1 filter stands out in relation to the others, especially for trend-following strategies in daily and weekly periods. In general, when you include costs in strategies, the filters present results that are higher than the benchmark, that is, unnecessary trades, thus reducing transaction costs. In this way, the proposed methodology is expected to provide support for decision-making by investors.