A technique is presented to increase the signal-to-noise ratio (SNR) in two-dimensional (2D), phase-encoded imaging at low SNR. The essence of this technique is to combine multiple echoes in the time domain. As analyzed in the paper, phase discrepancies exist among different echoes and may deteriorate the combined echo. In particular, extraneous phase shifts can be created if unshielded gradient coils are used. To overcome these phase discrepancies, a matched filter was derived from the k = 0 component of image. This matched filter has the same phase discrepancies among its echoes as the imaging signal and its magnitude decays with an average T2. In the echo summation with the matched filter, the phase of the matched filter was subtracted from the imaging signal and the magnitude of the matched filter was used as the weighting function. We have shown that this matched filter echo summation technique has better SNR than the case of 2D, phase-encoded imaging in both simulation and experiment. The SNR improvement is up to 60% in a phantom experiment. This technique is mostly useful in low SNR imaging that requires long imaging time, such as spectroscopic imaging and 19F imaging.