Several blind calibration methods have been proposed in a compressive sensing framework to mitigate the detrimental effects of uncertainties in the measurement matrix due to sensor gain and phase errors. Most of these methods operate on the signal domain samples of the receiving elements. This becomes computationally intractable if a large number of time samples is required, for example in low-SNR applications. In this paper, we propose an iterative blind calibration method to estimate the receiver path gains and phases as well as the observed scene from the measured array covariance matrix under the assumption that the observed scene is sparse. We successfully demonstrate the effectiveness of our method using simulated data for a 20-element uniform linear array as well as actual data from a 48-element station (subarray) of the Low Frequency Array (LOFAR) radio astronomical phased array.