We present MUSIC, a signal processing approach for identification of enriched regions in ChIP-Seq data, available atmusic.gersteinlab.org. MUSIC first filters the ChIP-Seq read-depth signal for systematic noise from non-uniformmappability, which fragments enriched regions. Then it performs a multiscale decomposition, using median filtering, identifying enriched regions at multiple length scales. This is useful given the wide range of scales probed in ChIP-Seq assays. MUSIC performs favorably in terms of accuracy and reproducibility compared with other methods.In particular, analysis of RNA polymerase II data reveals a clear distinction between the stalled and elongating forms of the polymerase.