This paper addresses the problem of radio frequency interference (RFI) in high-frequency (HF) radar. An efficient scheme is proposed for RFI suppression by designing receive filter in fast time domain for pulse compression. Throughout the implementation in observation data, this scheme does not need any training data. Instead, it uses the temporal data which is recovered after sea-clutter mitigation from observation data. As we know, sea-clutter is hard to eliminate completely. Residual clutter would contaminate the RFI covariance estimate and greatly degrade the performance of whitening filter. Considering the robustness to residual clutter, in this paper, filter design employs the similarity constraint which restricts the sought-after filter to be similar to the transmit signal. Under the criteria of maximum signal-to-interference-plus-noise ratio (SINR), the constrained optimal filter is analyzed, given in closedform. Experimental results demonstrate that the scheme works efficiently in the presence of residual clutter and reveals the target which is submerged by RFI in conventional processing.