Recently a number of high-resolution techniques for event detection and parameter estimation (velocity) have emerged in the field of multichannel sonar array processing. We find that a practical and computationally efficient implementation of these methods is possible for seismic reflection data. This implementation results in development of the covariance measure for event detection and velocity estimation in seismic reflection data. The measure is based on the eigenstructure of the sampled data covariance matrix. This decomposition is carried out within hyperbolic windows that are moved through common-depth-point (CDP) data. Eigenvalues of the data covariance matrix allow simultaneous estimation of the noise and signal energy present within each window, resulting in the development of a coherency measure.Computer implementation of the covariance measure provides a high resolution method for determination of velocity spectra relative to currently used techniques. The method applied to synthetic data resolves reflections closely spaced in time and velocity as well as those with low signal-to-noise ratio (S/N). These results are also achieved when the method is applied to real data. The results in each case are directly compared to the widely used semblance measure. The covariance measure provides results superior to the crosscorrelation-based semblance measure and offers a degree of resolution not previously attainable. These results can be achieved with a reduction in computational cost relative to semblance for typical analysis parameters. We find that the ability to continuously estimate noise energy, as well as signal energy, is critical to event resolution and noise rejection.