The wavelet-matched filter performs the continuous wavelet transform for edge feature enhancement and the correlation between two wavelet transforms for pattern recognition in a single step. The composite wavelet-matched filter is a combination of the wavelet-matched filters, that yields desired outputs for a given set of training mages. We present the definition, the iterative design algorithm, the quasi-orthogonality of the components, and the optical implementation of the composite wavelet-matched filter. The filter yields a sharp correlation peak at the origin without sidelobe. A comparison between the composite wavelet-matched filter with the conventional composite filter is also given.