The ability of marine vibrators to accurately control the frequency and phase of the emitted signal offers new and interesting possibilities. In terms of deblending, one could, for example, imagine having simultaneously operating vibrators in narrow non-overlapping frequency bands. Deblending, could then be done by applying a simple bandpass filter. In a sensitive survey area, one could imagine that vibrators omit the frequencies used by the local mammal population to communicate, thus reducing the overall environmental impact. In such cases, there is a need to interpolate or fill in the missing frequencies. In seismic processing, interpolating missing frequencies is a new problem, not directly related to the more well studied problem of interpolating missing spatial data. In this work, we present both classical signal processing methodology as well as CNN-based approaches for interpolation of missing frequency bands in seismic reflection data.@en