The Capon method is a powerful nonparametric approach in array processing based on the sample covariance matrix. For small sample sets, however, its performance is degraded. In this paper we formulate a regularized covariance matching framework based on the nuclear norm for enhancing the Capon method. An approximate iterative solution is developed and tested using simulated data. Appropriate regularization parameter values are also inferred from the data, drawing upon the cross-validation approach. The results show significantly improved spatial spectral and signal waveform estimates.