Among the many denoising methods developed in recent years, the FX filter is one of the most powerful algorithms that is used in daily seismic data processing. The conventional FX filter is actually a convolution filter with its convolution operator generated by an AR model. While reducing noise, this convolution operator can also smooth out some of the detailed information embedded in seismic data to an extent that depends on the operator length. In this paper, a new algorithm for an FX plus Deconvolution filter is proposed. The deconvolution step in this algorithm can recover the smeared signal which is generated by a conventional FX filter.