The characteristics of radio frequency interference (RFI) are analyzed for high-frequency surface wave radar (HFSWR), which adopts a linear frequency-modulated interrupted continuous wave. If RFI enters the receiver, it contaminates the positive and negative frequency range bins simultaneously, whereas the negative frequency range bins contain the RFI and noise only. Moreover, RFI is always spatially structured over a short duration. Based on the above characteristics, a new orthogonal projection filtering algorithm is proposed for RFI cancelation. The proposed method uses the information of antenna channel and sweep dimensions in negative frequency range bins to estimate the interference covariance matrix. To suppress RFI, the signals of interest in the positive range bins are projected onto the interference subspace, which is obtained by eigendecomposition of the covariance matrix. Experimental results suggest that the proposed method can achieve effective RFI cancelation by using data collected by a multifrequency HFSWR.