The problem of mismatch between repeated time-lapse seismic surveys remains a challenge, particularly for land acquisition. In this dissertation, we present a new algorithm, which is an extension of the surface-consistent model, and which minimizes the mismatch between surveys, hence improving repeatability. We introduce the concept of surface-consistent matching filters (SCMF) for processing time-lapse seismic data, where matching filters are convolutional filters that minimize the sum-squared error between two signals. Since in the Fourier domain, a matching filter is the spectral ratio of the two signals, we extend the well known surface-consistent hypothesis such that the data term is a trace-by-trace spectral ratio of two datasets instead of only one (i.e. surface-consistent deconvolution). To avoid unstable division of spectra, we compute the spectral ratios in the time domain by first designing trace-sequential, least-squares matching filters, then Fourier transforming them. A subsequent least-squares solution then factors the trace-sequential matching filters into four operators: two surface-consistent (source and receiver), and two subsurface-consistent (offset and midpoint). We apply the algorithm to two datasets: a synthetic time-lapse model and field data from a CO2 monitoring site in Northern Alberta. In addition, two common time-lapse processing schemes (independent processing and simultaneous processing) are compared. We present a modification of the simultaneous processing scheme as a direct result of applying the new SCMF algorithm. The results of applying the SCMF together with the new modified simultaneous processing flow reveal the potential benefit of the method, however some challenges remain, specifically in the presence of random noise.