This paper is concerned with differences in the frequency content of signal and noise on seismic traces. In order to develop a filter which has applicability over some considerable spatial range, special consideration is given to basic differences in the shape and the frequency content of individual signal and noise wavelets (events) on these traces. Therefore, a so-called “wavelet” Wiener filter is introduced which suppresses “noise” wavelets and enhances “signal” wavelets; this filter can be contrasted with an ordinary Wiener filter which discriminates between signal and noise on the additional basis of the statistics of the repetition of wavelets along the seismic trace. A technique for the automatic derivation of (wavelet) Wiener filters for seismic data by a digital computer is developed. The filters are time dependent to the extent that independent filters are derived at a sequence of data windows which are specified by the operator; criteria for selecting the window positions are given. The overall technique for the computer derivation of Wiener filters is demonstrated with synthetic and actual seismic data. A discussion of the wavelet Wiener filter and its relation to the ordinary Wiener filter is appended to this paper with a discussion of the effects of finite data windows on the eduction of the wavelet Wiener filter.